4.1. Burst Literature Exploration
Through the exploration of Citespace software, the preliminary result shows that the time-span of forming a non-null network in the original literature dataset is 1997–2019, with 3027 corresponding literature records in the analytic dataset. In this study, Citespace software is used to identify and summarize the articles with the top 20 burst strength, and Table 3
presents the descriptions of the 20 articles from a total of 67. The time interval of these 20 articles is roughly 2000–2010. In Citespace [14
], the burst literature means that the cited frequency of the literature has a jumping trend in a period, and the burst strength is a quantitative index measuring the rate of citation surge for the literature. The burst trend can be detected based on the Kleinberg’s algorithm [15
], which helps to find the core literature as pioneer of a new research area [16
The 20 articles that have burst strength in Table 3
generally reflect the main knowledge in the STI research field. From the burst literature, bibliometric information can be summarized for further research, including research background, key research issues, core theoretical models, and mainstream research methods. As seen in Table 3
, studies have formed a theoretical research basis, such as the research of Geels et al. [17
]. Moreover, the study of Smith et al. [21
] having the highest burst strength, mainly analyzed the quasi-evolutionary model of social technological system and expounded the selected situations during the transformation process of the system. Bergek et al. [22
] having the longest burst duration, mainly studied the functional dynamic evolution process of the innovation system and provided a practical reference for policymaking. The earliest published article is by Unruh et al. [23
], which indicated that innovation in sustainable technologies is mainly caused by fossil-fuel-based energy pathway dependence, and the dependence scale drives the co-evolution of technology and institutions.
Based on the historical background of energy transformation, the above studies regarded the development process of energy technology as a grand and complex system and put forward the theoretical framework of sociotechnical system, which focuses on the relevant mechanisms in the process of technology transformation. This theoretical framework is integrated with the technology evolution from a systematic perspective, and subsequent studies further formed new research perspectives, such as multilevel framework [24
], radical innovation [25
], and strategic niche [26
]. The knowledge base literature emphasizes that technology innovation in energy industry is closely related to the level of social and economic development and forms a development system similar to an ecosystem. In addition, it can be seen that the energy transformation starts from the sustainable development demand of the society, then forms the sustainable innovation of new energy technology, and finally develops the competitive advantage of strategic emerging industries.
4.2. Literature Co-Citation Analysis
Co-citation network of literature is constructed by applying Citespace, as shown in Figure 4
. To visualize the important nodes in the limited space, the software provides the g-index criterion to select nodes. The criterion means that the citation frequency of the selected node should be less than the square of citation ranking in all nodes, represented by g-index. Through the analysis, the literature co-citation network in the STI research field is obtained. The result shows that the literature co-citation network has a total of 1056 nodes, 3662 links, and the network density is 0.0066. The network density is the ratio of actual connections to potential connections, used as an index to describe the occupation of connections between nodes in the network. Clustering analysis can be carried out by combining the literature text content corresponding to the literature nodes.
reflects the theme relationships of the STI literature. The red text represents the cluster number and cluster label, and the corresponding shaded area is the literature nodes corresponding to the cluster category. The clustering result also shows the core co-citation network divided into 13 clustering categories, among which the largest five are “#0 Green Economy”, “#1 Business Model”, “#2 Green Innovation”, “#3 Regional Studies”, and “#4 Smart Cities”. Through transforming the perspective of bibliometric analysis, the literature co-citation network with cluster labels can be visualized in terms of timeline, as shown in Figure 5
. Compared with the cluster diagram in Figure 4
, the timeline diagram mainly shows the time interval distribution of the co-citation relationship in each cluster category. For example, the co-citation time interval of cluster category #0 is 2006–2017, while the co-citation time interval of cluster category #3 is 1999–2012.
It is seen from the cluster diagram that the network size of the other eight cluster categories is significantly smaller than that of the first five main categories. In the eight categories of the clustering, burst exploration analysis results indicate a small amount of burst literature. For example, Horbach [27
] investigated the related influence factors of environmental innovation based on Porter’s hypothesis and argued that environmental regulation and organizational change promote environmental innovation. This is the only one burst literature in the eight clustering categories, which is subordinate to the clustering category “#5 Moderating Factor” and has a burst strength of 6.6571. Apart from all 13 clustering categories, the burst exploration result shows that there are also a small number of burst articles in the non-mentioned cluster categories. The articles outside the 13 cluster categories reflect underlying research themes which can be the knowledge base of subsequent studies. For example, the studies of Wustenhagen et al. [28
] and Chesbrough [29
] have burst strengths of 3.4142 and 4.4238, respectively. The previous study focused on technological innovation in renewable energy, examining the impact of social acceptance on technological innovation. Additionally, the following research mainly discussed an issue related to the open innovation mode, proving how the business mode takes full advantage of the information and how the world-renowned companies operate the intellectual property rights.
The clustering results show five main clustering categories. The silhouette value [30
] is used to evaluate the literature clustering effect in this study, and the value close to 1 indicates the nodes having an ideal clustering result. “#0 Green Economy” is the largest clustering category in the literature co-citation network and the silhouette value of this cluster is 0.863. The network of cluster #0 contains 124 literature nodes, and there are 17 nodes presenting burst strength. Truffer et al.’s [31
] paper is the most actively cited in cluster #0. It explored the sustainable transformation process from innovation perspective and examined the spatial factors in a regional innovation system. The second-largest clustering category is “#1 Business Model” with a silhouette value of 0.896, and there are 113 literature nodes including 15 burst nodes in the cluster. The article of Boons et al. [32
] is the most active citer literature in cluster #1, which discussed the sustainable innovation issue and summarized the theoretical framework of sustainable business model. “#2 Green Innovation” is the third largest clustering category with a silhouette value of 0.854, and the cluster contains 103 literature nodes and 10 burst literature nodes. The study of Kiefer et al. [33
] is the most active citer literature in cluster #2. The paper identified the influencing factors related to ecological innovation and evaluated the promoting and hindering effects of the factors examined. The silhouette value of the clustering category “#3 Regional Studies” is 0.896. Cluster #3 is the fourth clustering category in the literature co-citation network containing 93 literature nodes and 18 burst nodes. The article of Alkemade et al. [5
] is the most active citer literature in cluster #3. It discussed sustainable innovation and argued that the competition challenge of emerging technology comes from the social expectation rather than technological performance. “#4 Smart Cities” is the fifth clustering category in the literature co-citation network. There are 39 literature nodes and four burst nodes in cluster #4, and the silhouette value of this cluster is 0.975. The research of Mora et al. [34
] is the most active citer literature in cluster #4. It mainly discussed the sustainable innovation issue on the smart city construction and summarized the strategic principles for promoting smart city development through four European city case studies.
Through the analysis of burst literature exploration and co-citation analysis with cluster, this paper summarizes the knowledge base and topic hotspots in the STI research field. The results indicate that subsequent research will further explore new research directions from the prior studies. With the deepening of the research process, research accumulation can lead to the transformation and evolution of the research topics. Therefore, according to the burst year rather than the publication year, this paper roughly divides the topic development of the STI research into four periods, which describe the evolution process of knowledge base and research topic hotspots. They are 2007–2009, 2010–2012, 2013–2015, and 2016–2017.
At the first stage (2007–2009), the corresponding research promoted the research process from a systematic perspective, such as energy technology system [35
], innovation system [36
], and energy economy system [37
]. Most literature nodes of this stage were subject to cluster #3. At the second stage (2010–2012), the multilevel theoretical framework emerged. The corresponding research aimed to explain and support sustainable technological innovation. Studies explored the application of the multilevel framework in an energy transformation process, such as technology change policy assessment [38
], energy transformation management [39
], and impact factors on the environment innovation [40
]. More literature nodes belonged to cluster #0, and a small number of literature nodes distributed in the cluster #2. At the third stage (2013–2015), researchers took interest in the public policy of sustainable innovation. More research contributed to the concept system of sustainable innovation, including grassroots innovation [41
], strategic niche evolution [42
], and business model [43
] among others. Relevant research articles were mainly distributed in cluster #0 and cluster #1. The research in the fourth stage (2016–2017) tended towards diversification, with research topics related to sustainable innovation further evolving to risk minimization of sustainable innovation [44
], diversity factors of sustainable innovation [45
], and innovation system framework [46
]. In addition, there were some emerging research topics such as responsible innovation [47
], green supply chain [48
], and smart-city construction [49
]. At this stage, many related articles distributed in cluster #2, while others were evenly distributed in cluster #0, cluster #1, and cluster #4.
Based on the above co-citation analysis, the early STI studies focused on research topics related to energy system transformation, mainly because the crisis of fossil fuel use then attracted wide attention from the international community. Besides, the concept of sustainable innovation at that time was not fully developed, and public policy was urgently needed to guide the transformation of the energy system [50
]. Subsequently, research topics were formed related to sustainable innovation, and STI research made significant progress from the theory perspective [51
]. In the recent progress of the STI research, the emergence of topics, such as smart cities [52
], indicated that the basic theories are relatively mature, and there was more attention paid the impact of emerging technologies on sustainability.
4.3. Main Path Analysis
Main path analysis can reflect the mainstream and relevant literature in the overall citation network, as well as highly cited papers in specific research fields [53
]. The arc path in the visualization results reflects the evolution of the research topic and methodology. In this study, we applied Pajek to perform the main path analysis, and visualized the result through VOSviewer. As shown in Figure 6
, the main path network has a total of 82 nodes and 120 directed arcs, and the density of the overall network is 0.018. The arc linking nodes has a direction. In addition, there are 12 layers in the main path network. The bottom layer is the sink vertex containing one paper, whereas the top layer is the source vertexes containing three papers in total. In a non-cyclic network, the sink vertex represents the node with an output degree value of 0, having no outgoing arcs to other nodes. The source vertex represents the node with an input degree value of 0, having no ingoing arcs to other nodes. For the literature node of sink vertex, Kemp et al. [54
] proposed a theoretical framework for strategic niche management of emerging technologies, which aims to realize the transition to a sustainable technology innovation mechanism. Besides, the literature node is the only sink vertex in the main path network, reflecting that this literature is the core knowledge base in the STI research field.
According to the attribute value ranking of arcs, the first 30 arcs in the main path network are presented in Table 4
. The attribute value is the traversal weight of citation, referring to the proportion of citation arc or literature node in all paths between source vertexes and sink vertexes. Besides, the attribute value reflects the dependence of the knowledge diffusion process on the citation arc or literature node in a specific discipline. As seen from Figure 6
and Table 4
, literature nodes related to research teams of Boons, Kemp, Coenen, Ritala, Hekkert, Foxon, or Quist demonstrate a more significant network centrality than other nodes in the main path network. Network centrality indicates the association situation of a specific node, which will be more important if the node has more links with others. We found two papers published by Boons et al. [7
] occupy the most core positions from the main path network, playing a transition role in the STI research and expanding the emerging research field of sustainable business model.
The knowledge diffusion venation of the STI research field can be further identified from the main path analysis results. As seen in Figure 6
, 10 knowledge diffusion branches are derived from the sink vertex “86 Kemp R, 1998”, which further spread and eventually connect with the source vertex literature. For the corresponding literature of three source vertexes, Fiorentino (2020) et al. [55
] and Massaro et al. [56
] analyzed the impact of intelligent technology and blockchain technology on the sustainable business model. Ljovkina et al. [57
] studied innovative resource management from the ethical perspective of sustainable development. Moreover, these articles cited a single prior literature respectively, including Peralta et al. [58
], Dentchev et al. [59
], and Calabrese et al. [60
], which are the only way to connect the three source vertexes in the main path network. Due to the limitation space, other knowledge diffusion paths are not reported in this section.