Renewable and Sustainable Energy Reviews 55 (2016) 361–370
Contents lists available at ScienceDirect
Renewable and Sustainable Energy Reviews journal homepage: www.elsevier.com/locate/rser
Structural properties and inter-organizational knowledge flows of patent citation network: The case of organic solar cells Hochull Choe a, Duk Hee Lee b,n, Hee Dae Kim c, Il Won Seo d a Policy Development Team, Strategy and Cooperation Division, Korea Research Institute of Chemical Technology (KRICT), 141 Gajeongro, Yuseong, Daejeon 305-343, Republic of Korea b Department of Business and Technology Management, KAIST, 335 Gwahakro, Yuseong, Daejeon 305-701, Republic of Korea c Future Strategy Team, Daegu Digital Industry Promotion Agency, 2139-12, Nam-gu, Daegu 705-701, Republic of Korea d Center for Technology Transfer, Korea Research Institute of Standards and Science (KRISS), 267 Gajeongro, Yuseong, Daejeon 305-340, Republic of Korea
art ic l e i nf o
a b s t r a c t
Article history: Received 16 February 2014 Received in revised form 24 July 2015 Accepted 27 October 2015
This paper identifies the structural properties of a technological knowledge network and the role of major organizations in the network, and analyzes actual contents of technological knowledge flows in terms of organization-technology linkage, by targeting the field of organic solar cells (OSC). Network analysis and matrix analysis methods are used to achieve these purposes. The results show the smallworld effect exists in the technological knowledge network of OSCs, and the organizations with high betweenness centrality lead the technological knowledge flows. We also find technological knowledge in classes 136 and 313 flows relatively actively in key organizations' network of OSCs. This means that technological knowledge regarding photoelectric batteries and electric lamp and discharge devices is mainly circulated between key organizations and indicates that the electronics or display sector will become a major consumer for early commercialization of OSCs. The target of analysis in this study is a patent citation network in the field of OSCs. Since we did not analyze all scientific publications, we cannot conclude that the results represent the entire flow of technological knowledge in that field. However, given that little attention has been paid to empirical studies of technological knowledge flows at the organizational level, this study makes an academic contribution by directly analyzing technological knowledge flows between organizations and presenting new taxonomic method based on centralities. The analytical process and methodology of this study, which include analysis of the structural properties of technological knowledge networks, matrix analysis and taxonomical grouping, and analysis of technological knowledge flows between key organizations, will be usefully applied to the analysis of technological knowledge networks in other fields. & 2015 Elsevier Ltd. All rights reserved.
Keywords: Network analysis Patent citation network Technological knowledge flow Organic solar cell
Contents 1. 2.
3.
4. 5.
n
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Background studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Solar photovoltaic industry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Organic solar cells: overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Patent data as a proxy for technological knowledge flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Network approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Research framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Measures for network analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data collection and network construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Corresponding author. Tel.: þ 82 42 350 6306; fax: þ 82 42 350 6831. E-mail addresses:
[email protected],
[email protected] (D.H. Lee).
http://dx.doi.org/10.1016/j.rser.2015.10.150 1364-0321/& 2015 Elsevier Ltd. All rights reserved.
362 362 362 363 363 364 365 365 365 365 365 365
362
H. Choe et al. / Renewable and Sustainable Energy Reviews 55 (2016) 361–370
5.2. Topological structure and node centrality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3. Structural properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4. Matrix analysis and taxonomical grouping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5. Technological knowledge flows between key organizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. Conclusions and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1. Introduction As the conventional economy that mainly depended on the inputs of labor and capital has been advanced to a knowledge-based economy, technological knowledge has been recognized as a driver of economic growth and even as a source of social change [1]. In this context, the creation and diffusion of technological knowledge has been accepted as a key factor that determines a country’s economic growth at the macro-level perspective as well as deciding a company’s competitiveness at the micro-level perspective. With growing recognition in recent years that the diffusion of technological knowledge triggers innovations between industries and provides a source of new growth, increasing research attention is being given to technological knowledge flows. Previous studies on technological knowledge flows are divided into three levels according to the subject of those flows: country, industry, and organization. Most of the preceding studies have focused on technological knowledge flows between countries or between industries. Studies at the country level mainly deal with the transfer patterns of disembodied knowledge between countries and the differences thereof by using patent citation data. Tseng [2] found, by analyzing knowledge networks of the information technology field between Asian countries, that while South Korea and Taiwan have created a number of gradual innovations in applied fields, radical innovations in the basic fields have mainly occurred in India and Singapore. Hu and Jaffe [3] studied the pattern of technological knowledge transfer from the U.S. and Japan to South Korea and Taiwan, and examined the differences between South Korea and Taiwan in the process of technological knowledge absorption. Jaffe and Trajtenberg [4] analyzed knowledge flows using patent citation data on five major assignee countries in the U.S. Patent and Trademark Office (USPTO), and demonstrated that there are specific citation patterns by country. Studies at the industry level generally divide an industry into several sectors and investigate technological knowledge flows between those sectors. For the studies at the industry level, patent citation data are also mainly used for empirical analysis. Park et al. [5] divided South Korea’s manufacturing industry into six clusters and checked the knowledge flows between clusters using the conventional patent citation indices and network indicators. Kim and Park [1] looked into the structural changes of technological knowledge between South Korea's industry sectors from the dynamic perspective based on network theory during the 1980s and the mid-1990s. Han and Park [6] proposed an exploratory method to measure knowledge flows between industries, which is based on patent analysis and input–output analysis. By using this method, they examined knowledge flows in traditional industries and emerging industries in South Korea from 1990 to 2000. The studies that looked into technological knowledge flows between organizations checked the patterns of knowledge sharing or the collaboration network structure between firms, usually using survey data or reference materials regarding strategic alliances [7,8]. Because they were not considered as the diffusion process of technological knowledge, the technological knowledge flows were not directly analyzed in these studies. Thus, these
366 366 367 368 369 369 370
studies have limitations in understanding the structure and actual contents of technological knowledge flows. In this paper, we consider technology diffusion as technological knowledge flows between individual organizations. This means that the transfer of technological knowledge from one organization to another is regarded as the flow of technological knowledge. Based on this background, the purposes of this paper include the following:
Creating a technological knowledge network based on patent data at the organizational level;
Identifying the structural properties of the network and roles of individual organizations in the network;
Analyzing the actual contents of technological knowledge flows in terms of organization-technology linkage. This paper targets organic solar cells (OSC). OSCs are receiving attention as the next generation energy source because of their ease of processing, mechanical flexibility, and the prospect of low manufacturing costs compared to conventional silicon solar cells [9,10]. Patent applications and registrations in the field of OSCs have been increasing rapidly since 2001 and, accordingly, the technological knowledge flow is very dynamic as well. Therefore, it is very timely that we look into the technological knowledge network in the field of OSCs at this point. This paper analyzes a technological knowledge network at organizational level, focusing on the following three points. First, we present the structural properties and their implications through the topological analysis and visualization of the network, and then we examine the importance and value of individual organizations through the centrality analysis. This is worthwhile in the sense that it provides a holistic view of the technological knowledge network in the field of OSCs and specific information about individual organizations. Secondly, this paper classifies the individual organizations in the network and figures out their roles in terms of technological knowledge flows through the matrix analysis. Third, we select key organizations in the network based on the results of the matrix analysis, and study technological knowledge flows between these key organizations using the U.S. Patent Classification (USPC). This is meaningful in that we can understand the actual contents of technological knowledge flows between organizations in the network.
2. Background studies 2.1. Solar photovoltaic industry The solar photovoltaic industry can be divided into an upstream manufacturing segment and a downstream installation segment. Fig. 1 shows the industry's supply chain. The upstream segment includes a manufacturing process from materials to modules for photovoltaic products, and the downstream segment includes system installation and operations [11].
H. Choe et al. / Renewable and Sustainable Energy Reviews 55 (2016) 361–370
363
Fig. 1. Supply chain of solar photovoltaic industry.
Fig. 2. Research flow chart.
The materials are divided into silicon, inorganic compounds and organic compounds. This field is technology-intensive and has a high entry barrier, which accordingly creates high added value. However, in descending order of the supply chain, namely solar cells, modules, system installation, operations and maintenance, markets become increasingly more competitive and laborintensive and have low added value. Since the energy conversion efficiency decides the price competitiveness of solar photovoltaic products, solar cells occupy a central position in the supply chain of the solar photovoltaic industry. According to the level of commercialization, solar cells can be classified as crystalline silicon solar cells (CSSC), thin-film solar cells, or OSCs. In the past 10 years, China has shown the most remarkable achievements in the solar photovoltaic industry. Since 2011, it has accounted for over 50% of global solar cell and module production [12]. However, this does not necessarily mean that China has the most advanced technologies, as it is still unable to secure technological competitiveness in core material sectors, and relies on imports of key raw materials and production equipment. Upstream markets, where technology skills are key success factors, have been led by western companies [13], which have also led the filing of the related patents. China is the main supplier of solar cells and modules but holds few patents. In fact, most of China's solar companies do business in only some sectors of the supply chain for the solar photovoltaic industry, such as solar cells or modules [14]. This phenomenon is also evident in the field of OSCs, on which this study focuses. The U.S., Japan and Europe have led the related research, and the filing of patents is dominated by those countries. Since OSCs are an emerging technology, only a handful of companies make products in the commercialization stage. Chinese companies have not yet distinguished themselves in this field. Patent share and market share of an industry are known to have a generally high correlation. However, it is not a usual case that the main producers are the driving forces of technology in the field of solar cells.
2.2. Organic solar cells: overview OSCs are classified largely into two types. One is dye-sensitized solar cells (DSC) based on light-absorbing pigment materials, and the other is organic thin-film solar cells (OTSC) based on organic semiconductor materials. DSCs produce electricity using photosynthetic processes where the dye molecules absorbed on a substrate, which consists of redox electrolyte, accept sunlight and create electrons [15]. Their energy conversion efficiency has reached 12% [16]. And with low manufacturing costs and ease of mass production, they have received attention in the research and industrial sectors as a new alternative to high-cost CSSCs [16,17]. OTSCs produce electricity by using indium-tin-oxide (ITO), a transparent electrode, as the anode, and a metallic material such as aluminum as the cathode, with an organic semiconductor as the material for the photoactive layer [18]. OTSCs have the advantages of the design flexibility of organic semiconductor materials and a variety of synthesis methods. Since OSCs such as DSCs and OTSCs are cheap, eco-friendly, and use easily supplied raw materials, they have very high future applicability and development possibilities. Despite the advantages mentioned above, OSCs face several technological challenges. OSCs are short-lived and have low efficiency compared with silicon-based solar cells. Also, as the unit area of OSCs increases, efficiency decreases. Therefore, OSCs are unlikely to be as efficient and stable as solar cells based on silicon [19]. In order to accelerate the commercialization of OSCs, their efficiency enhancement, long life and large scale are required. 2.3. Patent data as a proxy for technological knowledge flows There has been much debate on the limitations and usefulness of patent data as a proxy for technological knowledge flows. The very first limitation is regarding whether patents can represent technological knowledge. This stems from the obvious proposition that not all inventions are patented and patentable [20]. Secondly, the propensity to patent is different by industry or by firm [20,21]. In some industries, other tools such as trade secrets or trademarks may be used more often to protect technological know-how. Third,
364
H. Choe et al. / Renewable and Sustainable Energy Reviews 55 (2016) 361–370
Table 1 Measures for network topological analysis. Measures
Definition
Number of nodes Number of links Density Average degree
The total number of nodes in a network. The total number of links in a network. The ratio of actual links to all possible links in a network. The degree is the number of links that a node has to other nodes. The average degree is calculated by dividing the sum of all node degrees by the total number of nodes in a network. Number of components The component is an isolated sub-network in a network. The number of components indicates the number of independent groups in a network. Number of nodes in the largest component The total number of nodes in the largest component. Average path length The average value of the geodesic path length between any pair of nodes in a network. Diameter The length of the largest geodesic path in a network. Clustering coefficient A node's clustering coefficient is the ratio of the number of actual links between the node’s neighbors, to the number of the maximum possible links between those neighbors. The network’s clustering coefficient is the average of the clustering coefficients for all the nodes.
Table 2 Node centrality measures. Measures
Definition
Degree centrality
It is measured by using the sum of nodes directly connected to one node, so it has a meaning of local centrality. The degree centrality of Node i Pn aðN i ;N k Þ where aðN i ; N k Þ is 1 if, and only if, Node i (N i ) and Node k (N k ) are connected by (CðiÞd ) is calculated by the following equation: CðiÞd ¼ k ¼n1 1 a line, otherwise 0; n is the total number of nodes in a network. In the case of a directed network, we define two degree centralities, which are indegree and outdegree. It is measured by using the sum of the shortest path length between one node and all the other nodes. It has a meaning of global centrality including not only direct connections but indirect ones. The closeness centrality of Node i (C ðiÞc ) is calculated by the following equation: C ðiÞc ¼ Pn n 1 where dðN i ; N k Þ is the geodesic path length between Node i (N i ) and Node k (N k ); n is the total number of nodes in a network.
Closeness centrality
k ¼ 1
dðNi ;Nk Þ
In the case of a directed network, we define two separate closeness centralities, namely incloseness and outcloseness. Betweenness centrality It is used to measure the degree to which one node plays a role as a bridge or broker in a network. The betweenness centrality of Node i (C ðiÞb ) P gjk ðiÞ=gjk is calculated by the following equation: C ðiÞb ¼ ½ðn j o2Þkðn 1Þ=2 where gjk denotes the number of geodesics linking Node j and Node k; g jk ðiÞ) denotes the number of geodesics linking Node j and Node k (ja k) that pass through Node i; n is the total number of nodes in a network.
Table 3 Number of patents by assignee country.
Table 4 Number of patents by assignee organization.
Country
Number
Country
Number
Organization
Number Organization
Number
U.S. Japan Germany South Korea United Kingdom Switzerland
1978 820 238 90 79 57
Taiwan France Canada Netherlands Others Total
47 24 23 19 81 3456
Konarka Technologies Princeton University Boeing Semiconductor Energy Laboratory Eastman Kodak JX Crystals Sharp Canon
121 107 89 81
University of California IBM, Motorola Universal Display Samsung
51 45 44 42
Fuji Photo Film Sanyo Others Total
41 38 2516 3456
individual patents have very different technological and economic significance and value [22]. This is due to the fact that only a small number of patents have very high technological and economic value. Most patents have a very low value. These three aspects may reduce the value and the meaningfulness of patent data. However, despite these limitations to patent data, studies on knowledge flows using patent data have increased recently. The reason for this is that the value of a patent is generally proportional to the number of its forward citation counts [23–26], and patent citation flows provide information about the direction of technology diffusion [27]. Additionally, some studies empirically prove a significant correlation between backward citations and patent value [28–30] The biggest advantage of patent data is that they represent a direct outcome of the invention process [31]. Considering that inventions are the source of new technological knowledge, even though not all inventions are patented, patent data has very important significance as a criterion for technological knowledge or technological innovations. Furthermore, since patents provide
65 60 57 53
varied information accumulated over a long period of time, they are valuable as a proxy for technological knowledge flows. 2.4. Network approach Since patent citation data contain abundant information for analyzing disembodied knowledge flows between cited patents and citing patents, they have been one of the main indicators providing information on technological relationships among inventions [32]. Patent citation information has also been the most important and basic indicator to measure the impact of such patents [33]. The traditional methods used to understand technological knowledge flows using patents and patent citation data are the statistical analysis of the number of patents and the analysis of citation frequency [4]. However, since these methods have the limitations that they can provide only partial information on technological knowledge diffusion [34], further
H. Choe et al. / Renewable and Sustainable Energy Reviews 55 (2016) 361–370
Table 5 Number of patents by USPC class number.
365
Table 6 Results of network topological analysis.
Class number Title
Number
Measures
Measured value
136 428 257
1078 338 326
Number of nodes Number of links Density Average degree Number of components Number of nodes in the largest component (as a percentage) Average path length (average path length in a random network) Diameter (diameter in a random network) Clustering coefficient (clustering coefficient in a random network)
604 1331 0.004 2.119 5 594 (98.3%) 4.005 (7.398)
438 313 252 528 429 427 359 Others Total
Batteries: thermoelectric and photoelectric Stock material or miscellaneous articles Active solid-state devices (e.g., transistors, solid-state diodes) Semiconductor device manufacturing: process Electric lamp and discharge devices Compositions Synthetic resins or natural rubbers—part of the class 520 series Chemistry: electrical current producing apparatus, product, and process Coating processes Optical: systems and elements
274 245 127 69 68 66 52 813 3456
research has been carried out to examine the technological knowledge diffusion process from a more comprehensive perspective. In order to grasp the overall process of technological knowledge transfer or diffusion, those studies recognized the necessity to look into patent citation relations in terms of networks and analyzed patent citation data by combining them with network theory [27,32,34–37]. A network regards each actor as a node and denotes relationships between nodes as links. Network analysis is a quantitative technique to understand the interactive relationship among nodes or its structure. Network analysis methods are various and still evolving, but generally network topological analysis and node centrality analysis are used [37]. The former provides information about the entire structure of the network and a general perspective on relationships among nodes, while the latter helps to figure out the value and importance of individual nodes in the network [37]. Since the purpose of this paper is to study technological knowledge flows at the organizational level by using the patent citation network and to identify their implications, assignee organizations are represented by nodes and citations among assignee organizations are denoted as links which refer to the interactions among nodes.
3. Methodology 3.1. Research framework The research framework is shown in Fig. 2. The present study starts from collecting patents data in the field of OSCs. Next, we create the assignee organization citation network using backward citation and forward citation data from the collected patents. The analysis is carried out in the following three phases: network analysis, matrix analysis and taxonomical grouping, and analysis of technological knowledge flows between key organizations. 3.2. Measures for network analysis For analyzing the structure of a network, we use various statistical measures such as density, average degree, number of components, average path length, diameter, and clustering coefficient. These measures, suggested by Albert and Barabási [38], have been widely used in network literature for the purpose of understanding the structure and properties of the network [39,40]. Table 1 shows the measures and their definitions used in this paper for the network topological analysis. Next, we visualize the network to present an intuitive knowledge of its structure, and then carry out the centrality analysis using the measures suggested by Freeman [41] to examine the
9 (13) 0.319 (0.006)
value and significance of individual nodes. Table 2 shows the measures for estimating the centrality used in this paper.
4. Data collection and network construction This paper used the USPTO data to create the technological knowledge network in the field of OSCs. The USPTO database is well organized in terms of search conditions and reliability [42]. First, we made a query1 in order to extract registered patents in the field of OSCs, and by using the query, searched on title, abstract and exemplary claim. The search was performed on September 29, 2012, and showed that a total number of 219 patents were registered until 2011 since the very first patent was filed in 1977. Before 2000, fewer than five patents were registered each year, but from 2001 the number of registered patents began to increase rapidly. In 2011 only, 56 patents were registered during the year. It turned out that 186 patents comprising 84.9% of the total number were registered after 2001. To create the technological knowledge network, we analyzed the references displayed in 219 registered patents, and obtained data on 2222 backward and 1015 forward citations. This means that one registered patent has 10.1 backward citations and 4.6 forward citations on average. Finally, we created the assignee organization citation network in the field of OSCs by using the 219 registered patents and information on 3237 citations. In a patent citation network, backward citations indicate the degree of technology inflow, and forward citations show the inventive quality from the technological and economic perspective [43]. Therefore, we built it as a directed network.
5. Results 5.1. Descriptive statistics Table 3 shows the total number of 3456 patents, which consist of the technological knowledge network in the field of OSCs, by country. With 1978 patents, the U.S. holds 57.2% of the total number. The U.S. effectively leads the technological knowledge flows of OSCs, followed by Japan and Germany. Table 4 displays the number of patents by organization. Konarka Technologies ranks at the top with 121 patents, and Princeton University and Boeing follow with 107 and 89 patents respectively. The U.S. accounts for nine of the top 15 organizations, while five of the remaining six organizations are Japanese and one is Korean. By 1 The query was completed through consultation with researchers and a patent attorney engaged in relevant fields of OSCs at Korea Research Institute of Chemical Technology (KRICT).
366
H. Choe et al. / Renewable and Sustainable Energy Reviews 55 (2016) 361–370
Fig. 3. Assignee organization citation network.
type of organization, 13 out of 15 organizations are classified as private companies, and most of them are electronics companies or manufacturers of imaging and photographic products. Only Konarka Technologies and JX Crystals are solar energy companies. Table 5 shows the number of patents by USPC class number. 1078 patents belong to class 136, with the highest portion of 31.2%. The classes 428, 257, 438, and 313 form the secondary group. The number of patents in the top five classes is 2261, or 65.4% of the total patents. 5.2. Topological structure and node centrality This paper used the NetMiner 4.0 for analyzing the network. Table 6 shows the comprehensive analysis results of the entire network. Observing the basic structure, we find that the technological knowledge network of OSCs is composed of 604 nodes and 1331 links2, and forms a sparse connected network with a low density of 0.004. The average links per node is 2.119, which means that organizations in the network have relationships with around two other organizations on average. The entire network is divided into five component groups, and 594 nodes that cover 98.3% of all nodes belong to the largest component. Therefore, in fact, it is fair to consider it as a network composed of one component. Fig. 3 visualizes the entire structure of the technological knowledge network of OSCs. Nodes are indicated as dots and links as lines. The links are concentrated on nodes located in the center. Fig. 3 has the advantage that we can see the entire structure at a glance, but also has the disadvantage that it looks too complicated because so many nodes and links are indicated at once. In this case, we can visualize it 2 Number of links is fewer than 3237 (sum of backward citations and forward citations) because the weight (number of citations) is not reflected in the process of network creation. For cases in which more than one citing/cited data exist, we consider that an inflow/outflow link exists between the two nodes. This is called an unweighted network.
more clearly by using the k-core decomposition. The k-core decomposition is a method that identifies particular subsets called k-cores in the entire network, by recursively removing all nodes of a degree smaller than k until the degree of all remaining nodes is larger than or equal to k [40]. Fig. 4 shows 51 nodes included in the subset of 6-core. In Fig. 4, the thickness of a line is proportional to the frequency of citation, and the arrow shows the direction of citation. Visually the network is divided largely into two main clusters. Konarka Technologies and Universal Display are located in the center of each cluster. Sharp and Princeton University are the bridges to connect two main clusters, and IBM, Boeing, Eastman Kodak and École Polytechnique Fédérale de Lausanne (EPFL) play a role in connecting the clusters' core and periphery. These central organizations are private companies, except for some universities. This phenomenon is also identified from Table 7, which displays the top 10 organizations ranked by three centralities. Like the above k-core decomposition, most of the organizations are private companies, except for some universities. Among them, Konarka Technologies, Universal Display and Sharp occupy the most critical positions. 5.3. Structural properties Table 6 indicates that the average path length of the technological knowledge network of OSCs is 4.005. This means that each node is connected to other nodes through four steps, on average. This value is much smaller than 7.398, the theoretical predictive value of the random network proposed by Erdös and Rényi [44] (hereafter the ER model), which has the same number of nodes and links. In addition, the diameter, indicating the maximum value of the shortest path length between any pair of nodes, is 9. This value is also much smaller than 13, the theoretical predictive value of the ER model. These facts mean that the small-world effect is stronger in the technological knowledge network of OSCs than in the ER model, and imply that it has autonomously changed into
H. Choe et al. / Renewable and Sustainable Energy Reviews 55 (2016) 361–370
367
Quantum Group Isis Innovation
JX Crystals
Eastman Kodak Shipley
Polaroid
Nokia
Solaria
Allied Signal Forskarpatent I Uppsala AB
PolyIC & Co. KG Advanced Refractory Technologies
Ormecon Zipperling Kessler
Photon Power Loctite Luminescent Systems IBM
Nanosys Boeing
Nanosolar EPFL
UltraDots
University of California
Princeton University
Vitex Systems
Toshiba
Seagate Technology Sharp
Agilent Technologies
Konarka Technologies
Universal Display Osram Opto Semiconductors Gracel Display
Canon GE
Pioneer
Idemitsu Kosan
The Standard Oil
Semiconductor Energy Laboratory
Fuji Xerox United Solar Systems
Motorola
Battelle Memorial Institute
Fuji Photo Film
Sanyo
Atlantic Richfield
Energy Conversion Devices
Samsung Dieter Meissner
Finisar QSEL-Quantum Solar Energy
Fig. 4. Assignee organization citation 6-core network.
Table 7 Top 10 organizations with the highest centrality. Rank Degree centrality
Closeness centrality
Betweenness centrality
In
Out
In
1
Konarka Technologies (0.580)
Konarka Technologies (0.240) IBM (0.182)
2 3
Universal Display (0.151) Samsung (0.138)
Universal Display (0.194) EPFL (0.186) Sharp (0.177)
4
Ormecon (0.124)
5
Semiconductor Energy Laboratory (0.108) JX Crystals (0.096) Boeing (0.090) MIT, Princeton University (0.081) Toshiba (0.088) – Princeton University (0.081) Sharp, Nomadics (0.073) Canon (0.080) – Motorola (0.073)
6 7 8 9 10
Out
Samsung (0.227) Sharp (0.198)
Eastman Kodak (0.114) Eastman Kodak, Carlson, (0.197) IBM (0.095) – Nokia (0.196) KIST (0.192) Sumitomo Metal Mining (0.191) Princeton University (0.188) Alliance for Sustainable Energy (0.187)
the network effectively transferring the knowledge. Meanwhile, the clustering coefficient is 0.319. This value is much bigger than 0.006, the theoretically estimated value of the ER model of the same size. This conforms with a theory of Watts and Strogatz [45], who argue that many networks in the real world have much larger local clustering than that which appears in the random network. As a result, the technological knowledge network of OSCs has a shorter average path length and a higher clustering coefficient than the ER model of the same size. This fact means that the technological knowledge network of OSCs is neither entirely
Sharp (0.057)
Sharp (0.165) Eastman Kodak (0.047) Eastman Kodak, Exxon Research and Konarka Technologies (0.046) Engineering (0.164) – IBM (0.044) Boeing, Toshiba (0.162)
Boeing (0.031)
– EPFL (0.154) Motorola (0.152)
MIT (0.030) EPFL (0.025) KIST (0.023)
Licentia Patent (0.148) Polaroid (0.145)
Universal Display (0.021) Princeton University, Samsung (0.018)
random nor entirely well-organized as implied in the small-world network by Watts and Strogatz [45].
5.4. Matrix analysis and taxonomical grouping We position individual nodes on the matrix using two kinds of indices and then taxonomize them. In order to compare the degree of technological knowledge inflow (inflow link) with the inventive quality (outflow link) of each node, we developed the outflow–
368
H. Choe et al. / Renewable and Sustainable Energy Reviews 55 (2016) 361–370
Fig. 6. Technological knowledge flows in the 11 key organizations’ network (cut off: 3).
Fig. 5. Distribution of organizations with O–I index and betweenness centrality.
inflow (O–I) index first as the equation (1) below. O I index ¼
ðOutdegree centrality–Indegree centrality Þ ðOutdegree centrality þIndegree centrality Þ
ð1Þ
The O–I index is conceptually similar to the external–internal (E–I) index [46]. While the E–I index compares the number of internal links between members of a group with the number of external links to other group members to analyze the differences between those groups3, the O–I index compares the degree to which a node cites other nodes with the degree to which that a node is cited by other nodes using the concept of degree centrality at the individual node level. The O–I index has a value between 1 and 1. When the value is greater than 0, this indicates that the number of cited links is more than the number of citing links, and when the value is smaller than 0, it shows that the number of citing links is more than the number of cited links. In other words, the closer the O–I index of a node gets to 1, the higher its inventive quality becomes. Conversely, if the O–I index of a node is close to 1, this means that the influx of technological knowledge to the node is increasing. By using the O–I index, we can easily figure out whether a particular node is a knowledge producer or a knowledge absorber in the citation network. However, since the O–I index compares the number of inflow links with the number of outflow links at the individual node level, it cannot provide absolute information about the importance and role of a particular node in the whole network. This problem can be solved by using the betweenness centrality as another index. As explained in Table 2, betweenness centrality is defined as the number of geodesic paths that pass through a node. Nodes with high betweenness centrality play a role in bridging information flows between nodes in the whole network, and are generally called brokers or gatekeepers. The matrix analysis by using both the O–I index and the betweenness centrality provides an inductive taxonomy of nodes based on the knowledge flows. Through this analysis, we can examine the role of individual nodes in technological knowledge networks and the implications thereof. 3 Nodes can be grouped according to their roles, properties and structural locations in the network.
Fig. 5 represents the distribution of assignee organizations with the O–I index and betweenness centrality. As shown in Fig. 5, we can classify them into four categories of taxonomy based on the assignees' role in the network: (A) brokers based on producing knowledge, (B) brokers based on absorbing knowledge, (C) knowledge producers, and (D) knowledge absorbers. Each category has the following features: (A) the O–I index is more than 0 and betweenness centrality is relatively high, (B) the O–I index is less than 0 and betweenness centrality is relatively high, (C) the O–I index is more than 0 and betweenness centrality is relatively low, (D) the O–I index is less than 0 and betweenness centrality is relatively low. The organizations in group A play a role in creating patents with relatively high inventive attractiveness and bridge the flows of technological knowledge in the field of OSCs. The organizations in group B play a role as brokers for technological knowledge by actively absorbing technological knowledge in the field of OSCs. The organizations in group A or in group B are key nodes that are mostly located in the center of the network and have many links. Most organizations in the network belong to group C or group D. The organizations in Group C provide technological knowledge to the field of OSCs, but have low outdegree centrality and betweenness centrality. The organizations in group D absorb technological knowledge in the field of OSCs, but display low indegree centrality and betweenness centrality. The organizations in group C and group D are located on the periphery of the network. 5.5. Technological knowledge flows between key organizations Since the matrix analysis developed in this paper is based on node centrality, it provides information on the importance and value of individual nodes. However, the node centrality cannot provide information on the contents of technological knowledge flows between individual nodes. To understand the contents, we used USPC to analyze 11 key organizations in group A and group B, which lead the technological knowledge flows. Technological field classifications such as USPC or IPC are widely used in many studies that have analyzed technological knowledge flows or technological trajectories by using patent data [32,34,35,47]. This paper goes one step further from the previous studies by attempting the organization-technology linkage analysis. Technological knowledge flow analysis linking the organization (11 key organizations) and technological field classification (USPC) provides an opportunity to observe not only the direction and size but also the actual contents of technological knowledge flows. Fig. 6 represents the technological knowledge flows between 11 key organizations.
H. Choe et al. / Renewable and Sustainable Energy Reviews 55 (2016) 361–370
This figure indicates that the knowledge in class 136 or class 313 flows most actively among 11 key organizations, in the center of which Konarka Technologies and Universal Display are located respectively. Konarka Technologies absorbs the technological knowledge in class 136 mainly from Sharp, Princeton University and EPFL. Fig. 6 also indicates that Konarka Technologies has few self-citations and technology transfers. This implies either that although Konarka Technologies is actively absorbing technological knowledge, it has not yet succeeded in creating technologies with high inventive quality, or, considering that Konarka Technologies was established in 2001, that its patent applications were only relatively recently filed and there has not been sufficient time to evaluate the quality of those patents. Universal Display absorbs the technological knowledge in class 313 mostly from Sharp and IBM. Its active self-citations imply that the company makes efforts to internalize the knowledge to develop innovative Organic Light-Emitting Diode (OLED) technology for flat panel displays and electronics, which is its main field of business. Sharp and EPFL deserve attention from the perspective of technological knowledge supply. Sharp transfers its knowledge of class 136 to Konarka Technologies, while technological knowledge of class 313 is transferred to Universal Display. This means that Sharp is equipped with technologies with high inventive attractiveness in both batteries and optics, and at the same time explains well why Sharp has the highest betweenness centrality. Only EPFL supplies technological knowledge of class 429 to the 11 key organizations’ network. EPFL focuses on DSCs, and the results so far indicate that EPFL creates technological knowledge with very high inventive quality mainly by internal research and development (R&D). EPFL's knowledge of class 429 mainly flows into Konarka Technologies, while some of it flows into Samsung. This implies that Konarka Technologies and Samsung put effort into the development of DSCs as well. However, Konarka Technologies licensed out all the technologies in the field of DSCs in 2002 or so, and have concentrated on developing technologies of OTSCs since then. In the meantime, since first applying for a patent in the field of DSCs in 2003, Samsung has continuously filed patent applications in the field. Princeton University is a supplier and absorber of technological knowledge in the 11 key organizations' network and focuses on the field of OTSCs. Absorbing the technological knowledge of class 313 from Eastman Kodak, Princeton University combines it with internal R&D and creates the technological knowledge of class 136. The knowledge of class 136 by Princeton University mainly flows into Konarka Technologies. In the 11 key organizations' network, MIT supplies the technological knowledge of class 136 only to Boeing. Creating new knowledge of class 136 by combining MIT’s with internal R&D, Boeing then transfers it to Konarka Technologies. Eastman Kodak and IBM along with Sharp supply the technological knowledge of class 313 to the network. KIST does not exchange technological knowledge with other key organizations.
6. Conclusions and discussion In this paper, we analyzed the technological knowledge flows in the field of OSCs using patents and patent citation data, which are inevitably produced in the process of innovation. The main findings are as follows:
Technological knowledge in the field of OSCs is driven by the U. S. in terms of country, electronics companies and manufacturers of imaging and photographic products in terms of assignee, and class 136 in terms of the technology field;
369
The small-world effect exists in the network; The network consists of two main clusters, and the result of the
k-core decomposition generally conforms with that of centrality analysis; The nodes with high betweenness centrality are located in the center of the network and lead technological knowledge flows; In the network of 11 key organizations, technological knowledge flows in class 136 and class 313 are relatively active.
There is a paucity of empirical study on technological knowledge flows at the organizational level. Therefore, the present study makes an academic contribution by identifying the structural properties of technological knowledge networks and the actual contents of knowledge flows by analyzing the field of OSCs. In addition, we present a new taxonomic method. Based on the results of this study, the following implications are provided for the future direction of technology development in OSCs. Firstly, the direction of technology commercialization is discussed. Ameri et al. [9] reported that CSSCs occupy about 90% of the worldwide solar cell production. These first generation solar cells’ overwhelming dominance gives rise to negative opinions on the commercialization of OSCs. However, as the efficiency and stability of OSCs has dramatically improved recently, much interest is focusing on the feasibility of their commercialization. We found that out of the top 15 organizations in the network, 10 organizations are electronics companies or producers of imaging and photographic equipment (see Table 4). In addition, the results of organization-technology linkage analysis among 11 key organizations reveal that the technologies in class 136 mainly flow into the solar energy company, while those in class 313 are generally brought into the developer of OLED technology. This suggests that early commercialization of OSCs should be directed to facilitate device-integrated photovoltaics (DIPV) such as for-electronicdevices or disposable batteries, rather than for the use of mass production of electricity requiring high efficiency. Thus, the strategic direction of commercialization should be to develop lowprice flexible solar cells with multiple applications. This strategy is to avoid competing with CSSCs in the market. Secondly, from the viewpoint of the industrial ecosystem, there is a need for active linkage between OSC producers and electronics and display companies. As mentioned earlier, most of the demand for initial commercialization of OSCs will be created by the electronic and display sector. Therefore, OSC producers need to set up technology development strategies based on the needs of the demand sector. In this sense, it is very important to secure cuttingedge technologies customized for the development of products for the demand companies as well as to create partnerships with them. This study analyzes the network created by using patents and patent citation data in a particular field of OSCs. Since not all scientific publications on OSCs are covered as a target for the analysis, it cannot be concluded that the results of this study represent the entirety of the technological knowledge flows in OSCs. Also, since the patent data used in this study was extracted from USPTO, it may not be a sample that is free of bias toward one particular country, the U.S. Despite these limitations, this study serves as a good analysis model to find the implications of technological knowledge networks. The methodology used in this study can be applied to a technological knowledge network in other fields.
Acknowledgments This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (2014S1A3A2044459).
370
H. Choe et al. / Renewable and Sustainable Energy Reviews 55 (2016) 361–370
References [1] Kim MS, Park Y. The changing pattern of industrial technology linkage structure of Korea: Did the ICT industry play a role in the 1980s and 1990s? Technol Forecast Soc Change 2009;76:688–99. [2] Tseng CY. Technological innovation and knowledge network in Asia: evidence from comparison of information and communication technologies among six countries. Technol Forecast Soc Change 2009;76:654–63. [3] Hu AGZ, Jaffe AB. Patent citations and international knowledge flow: the cases of Korea and Taiwan. Int J Ind Organ 2003;21:849–80. [4] Jaffe AB, Trajtenberg M. International knowledge flows: evidence from patent citations. Econ Innov New Technol 1999;8:105–36. [5] Park J, Lee H, Park Y. Disembodied knowledge flows among industrial clusters: a patent analysis of the Korean manufacturing sector. Technol Soc 2009;31:73–84. [6] Han YJ, Park Y. Patent network analysis of inter-industrial knowledge flows: the case of Korea between traditional and emerging industries. World Pat Inf 2006;28:235–47. [7] Gupta AK, Govindarajan V. Knowledge flows within multinational corporations. Strateg Manag J 2000;21:473–96. [8] Verspagen B, Duysters G. The small worlds of strategic technology alliances. Technovation 2004;24:563–71. [9] Ameri T, Dennler G, Lungenschmied C, Brabec CJ. Organic tandem solar cells: a review. Energy Environ Sci 2009;2:347–63. [10] Hoppea H, Sariciftci NS. Organic solar cells: an overview. J Mater Res 2004;19:1924–45. [11] Barua P, Tawney L, Weischer L. Delivering on the clean energy economy: the role of policy in developing successful domestic solar and wind industries. World Resources Institute Working Paper; 2012. [12] Seo JH. The Trend and implications of Global Solar PV Industry. World Energy Market Insight. Seoul: Korea Energy Economics Institute; 2013. p. 13–42 (in Korean). [13] De La Tour A, Glachant M, Ménière Y. Innovation and international technology transfer: the case of the Chinese photovoltaic industry. Energy Policy 2011;39:761–70. [14] Joo D-Y, Seo DH, Kim JK. An analysis on competitive structure and cooperation between Korea and China in solar photovoltaic industry. Reserach reports, 2013, Seoul: Korea Institute for Industrial Economics and Trade, 2013-681 (in Korean). [15] Grätzel M. Dye-sensitized solar cells. J Photochem Photobiol C: Photochem Rev 2003;4:145–53. [16] Daeneke T, Kwon TH, Holmes AB, Duffy NW, Bach U, Spiccia L. High-efficiency dye-sensitized solar cells with ferrocene-based electrolytes. Nat Chem 2011;3:211–5. [17] Chiba Y, Islam A, Watanabe Y, Komiya R, Koide N, Han L. Dye-sensitized solar cells with conversion efficiency of 11.1%. Jpn J Appl Phys Part 2 Lett 2006;45:638. [18] Fujishima D, Kanno H, Kinoshita T, Maruyama E, Tanaka M, Shirakawa M, et al. Organic thin-film solar cell employing a novel electron-donor material. Sol Energy Mater Sol Cells 2009;93:1029–32. [19] Kumar P, Chand S. Recent progress and future aspects of organic solar cells. Prog Photovolt Res Appl 2012;20:377–415. [20] Archibugi D, Pianta M. Innovation surveys and patents as technology indicators: the state of the art. Innovation, patents and technological strategies. Paris: OECD; 1996. p. 17–56. [21] Magnusson T, Berggren C. Entering an era of ferment-radical vs incrementalist strategies in automotive power train development. Technol Anal Strateg Manag 2011;23:313–30. [22] Griliches Z. Patent statistics as economic indicators: a survey. J Econ Lit 1990;28:1661–707.
[23] Stolpe M. Determinants of knowledge diffusion as evidenced in patent data: the case of liquid crystal display technology. Res Policy 2002;31:1181–98. [24] Park G, Park Y. On the measurement of patent stock as knowledge indicators. Technol Forecast Soc Change 2006;73:793–812. [25] Yasukawa S, Kano S. Validating the usefulness of examiners' forward citations from the viewpoint of applicants' self-selection during the patent application procedure. Scientometrics 2013:1–15. [26] Harhoff D, Narin F, Scherer FM, Vopel K. Citation frequency and the value of patented inventions. Rev Econ Stat 1999;81:511–5. [27] Chang SB, Lai KK, Chang SM. Exploring technology diffusion and classification of business methods: using the patent citation network. Technol Forecast Soc Change 2009;76:107–17. [28] Bessen J. The value of US patents by owner and patent characteristics. Res Policy 2008;37:932–45. [29] Harhoff D, Scherer FM, Vopel K. Citations, family size, opposition and the value of patent rights. Res Policy 2003;32:1343–63. [30] Reitzig M. Improving patent valuations for management purposes—validating new indicators by analyzing application rationales. Res Policy 2004;33:939– 57. [31] Michie J. Introduction. The Internationalisation of the innovation process. Int J Econ Bus 1998;5:261–77. [32] No HJ, Park Y. Trajectory patterns of technology fusion: trend analysis and taxonomical grouping in nanobiotechnology. Technol Forecast Soc Change 2010;77:63–75. [33] Chen C, Hicks D. Tracing knowledge diffusion. Scientometrics 2004;59:199–211. [34] Li X, Chen H, Huang Z, Roco MC. Patent citation network in nanotechnology (1976–2004). J Nanopart Res 2007;9:337–52. [35] Huang Z, Chen H, Yip A, Ng G, Guo F, Chen ZK, et al. Longitudinal patent analysis for nanoscale science and engineering: country, institution and technology field. J Nanopart Res 2003;5:333–63. [36] Wartburg IV, Teichert T, Rost K. Inventive progress measured by multi-stage patent citation analysis. Res Policy 2005;34:1591–607. [37] Choe H, Lee DH, Seo IW, Kim HD. Patent citation network analysis for the domain of organic photovoltaic cells: country, institution, and technology field. Renew Sustain Energy Rev 2013;26:492–505. [38] Albert R, Barabási AL. Statistical mechanics of complex networks. Rev Mod Phys 2002;74:47–97. [39] Okamura K, Vonortas NS. European alliance and knowledge networks. Technol Anal Strateg Manag 2006;18:535–60. [40] Breschi S, Cusmano L. Unveiling the texture of a European Research Area: emergence of oligarchic networks under EU Framework Programmes. Int J Technol Manag 2004;27:747–72. [41] Freeman LC. Centrality in social networks conceptual clarification. Soc Netw 1979;1:215–39. [42] Lee C, Song B, Park Y. How to assess patent infringement risks: a semantic patent claim analysis using dependency relationships. Technol Anal Strateg Manag 2013;25:23–38. [43] Henderson R, Jaffe AB, Trajtenberg M. Universities as a source of commercial technology: a detailed analysis of university patenting, 1965–1988. Rev Econ Stat 1998;80:119–27. [44] Erdős P, Rényi A. On the evolution of random graphs. Publ Math Inst Hung Acad Sci 1960;5:17–61. [45] Watts D, Strogatz S. Collective dynamics of small-world networks. Nature 1998;393:440–2. [46] Krackhardt D, Stern RN. Informal networks and organizational crises: an experimental simulation. Soc Psychol Q 1988;51:123–40. [47] Hu MC. Knowledge flows and innovation capability: the patenting trajectory of Taiwan's thin film transistor-liquid crystal display industry. Technol Forecast Soc Change 2008;75:1423–38.