How does aggregation‐induced emission aggregate interdisciplinary research?

It is a matter of debate whether the discipline independence in discipline formation narrows its interdisciplinarity. It is also less well understood how disruptive works emerge in investigative practice rather than a theory‐driven approach. Aggregation‐induced emission (AIE) is an atypical photophysical phenomenon, in which the whole (aggregate) is brighter than the sum of its parts (single molecule). Through measuring and computing the cognitive extent and evolution of research on AIE, including topics, epistemic‐social collaborative networks, interdisciplinarity, emergent concepts, core concept networks and knowledge flow, this study shows that a cross‐research scales concept and its practice can establish new bridges in the sciences and promote disruptive work. Focusing on mesoscale entities, scientists from many different branches of science are involved in theoretical research on mechanisms, as well as developing different AIE systems for applications. The data analysis in this study provides details showing how non‐reductionist concepts based on new scientific discoveries cross traditional disciplinary boundaries and aggregate interdisciplinary research. The emergence and evolution of the AIE field implies that scientists may be motivated to embrace nonreductionist ideas at different research scales, leading to a more permeable field boundary.

However, much remains unknown about the investigative process of successful disruptive sciences, and it is less well understood how disruptive works emerge in investigative practice rather than a theory-driven approach.Investigating the progress of science using a different approach by analyzing scientific actions rather than products is beneficial for deepening the understanding of both the epistemic and social dimensions of disruptive works. [9]12][13][14] Some have suggested that new cross-disciplinary fields, [8] or a concept on doing science, such as click chemistry, [15] can establish new bridges in the sciences.Meanwhile, in the recent philosophy of chemistry, it is assumed that the prospects of discipline-and research-scale reduction are rather bleak, [16] as chemistry deals with substances with chemical microspecies-like molecules.However, these arguments have raised new questions regarding reduction, emergence, and physicalism, in which the directions or scales of scientific explanation lie.Is this explanation downwardsfrom the molecular structure (molecular scale) to the motions of the (microscope) parts, or upwards-from the molecular structure to the (macroscopic) materials in nature or produced in the laboratory or industry?How do scientists in cross-disciplinary fields manage entities of different scales that possess many structures and scales?
Aggregation-induced emission (AIE) is an atypical photophysical phenomenon, in which the whole (aggregate) is brighter than the sum of its parts (single molecule); it was first posited by Tang Ben Zhong and his team in 2001. [17]n the following 20 years, the field of AIE underwent expeditious and steep growth beyond chemistry to include multiple disciplines. [18]More importantly, aggregate science, a fast-growing interdisciplinary research frontier, is being developed in the AIE field to explore entities in a mesoscopic scale. [19]Taking the field of AIE as an example, we aim to discuss in what sense we may speak of a paradigm change, [20,21] and how disruptive work was invented along with new phenomena, concepts, tools, and ontological assumptions in the explanatory and investigative agendas of the scientific community.
To address the features and evolution of the AIE field, we combined computational methods and the perspective of the philosophy of science to measure the cognitive extent and evolution of the AIE field.Using the bibliometric analysis, the previous work gave a historical overview on the AIE field in detail.Besides the commonly used publication and citation metrics, [22] multiple data analysis methods were used in this article, such as topic models, interdisciplinarity, category normalized citation impact (CNCI) metric, and network analysis.These methods were ultilized to quantify and visualize the cognitive extent and evolution of the AIE field from different perspectives, including topics, epistemic-social collaborative networks, interdisciplinarity, emergent concepts, core concept networks, and knowledge flow.In other words, we perceive articulating and applying core theories as part of scientific practice but not the essence of scientific practice.The manipulating aspects of the science at different research scales are crucial for understanding disruptive works.

Corpus retrieval, bibliometric data, and cleaning
Two large domains of knowledge were comprehensively investigated: all AIE papers and cited papers from Web of Science (WoS) from 2001 to 2021.Considering a large amount of papers on AIE did not contain "aggregationinduced emission" directly but other similar phases or the description of typical mechanisms of AIE in their abstracts, titles, or keywords, the search queries were not set as "TS = aggregation-induced emission" so as to collect the accurate and complete information.((TS = ("Aggre* Induc* Emissio*" OR "Restrictio* of Intramolecu* Motio*")) and PY = (2001-2021) were used as search terms in the WoS firstly.The searched items were filtered individually to check for typographical errors, mistakes, or missing information in the original datasets.After the first-round search, "Aggre* enhanc*" Near/2 emissio*, "aggregation* induc*" Near/4 emissio*, "aggre* induc*" Near/3 phosphorescen*, "aggre* induc*" Near/2 enhanc* were added as search terms to make sure we get the complete and exact publications.Document types are restricted to "article" and "review", which pose a good representation of original research.Subsequently, the searched publications were compared with the documents in InCites for cross-validation.The misspellings of names, keywords, institutions, or countries in the original datasets are corrected, such as "ATE", "ATR" or "ME" are corrected to "AIE" and "shangdong univ" is corrected to "shandong univ".This procedure resulted in a set of 10671 publications on AIE and 410557 cited publications on AIE.
The raw data cleaning and processing were conducted using the software R. Gephi tool was used to visualize the spatialization process of collaboration and interdisciplinarity.Gephi is an efficient tool for network analysis used in social network analysis, biology, genomics, and many other disciplines.For network analysis, ForceAtlas2 is Gephi's default layout algorithm aimed at providing a generic way to spatialize networks, [23] which can convert all listed institutions, authors, and research fields into nodes and edges.

Synonymous keywords merging and institutions deduplication
To precisely measure the growth of the cognitive content of the publications on AIE, synonymous terms were merged and a dataset associating normalized synonymous terms was generated.In scientific papers, keywords are represented as terminology with specific meanings.Considering numerous synonymous scientific terms provide significantly different meanings regarding scientific concepts, methods, or techniques, a computer-implemented method was not used.However, synonymous terms were merged based on expertise in identifying their functions in scientific research.For example, "tetraphenylethylene moieties" and "tetraphenylethylene molecules" are merged into one, but "AIE molecules" and "AIE monomer" are retained as separate keywords.
Since the same institution usually exists in more than one type of abbreviation in WoS, the full names of the institutions were carefully checked and extracted.In the same institution, regardless of the number of times it appeared in the publication year, it was counted as only one type.Institutional deduplication was processed to measure the growth trends of institutions in the AIE field.Since one paper is usually collaboratively published with more than one institution, it was attributed to each institution when analyzing the trends of publications according to institutions.If two or more authors contribute to the same paper X and are affiliated with the same institution Y, Y is counted as one institution, and the number of papers attributed to Y is counted as one.

Data analysis: topic models, interdisciplinarity, CNCI, and network analysis
Scholars from the research fields of science of science and philosophy of science recognize that research topics, including entities, concepts, theories, tools, and methods, rise and fall in scientific interest and research areas, and that a certain degree of consensus is necessary for the progress of science. [24]The conceptual journey of the AIE field is investigated regarding infrastructure, research tools, and scientific methods using the topic model of latent Dirichlet allocation (LDA). [25]The LDA MALLET was used to generate and compute the topics.The coherence model was used to measure topic coherence to determine the number of topics required to best account for the structure of our corpus. [26]The emerging words were measured from the collected dataset.The reliability of each method was tested by the review papers of the AIE field as well as the consistency of each method's outcome.New computational approaches to unsupervised machine learning can scan an entire document to generate topics in a bottom-up approach based on the semantic structure of publications.Such algorithms make it possible to explore the semantic structure without any prior content-related knowledge, annotations, or labeling of documents and any preconceived hypotheses of the topics present.Therefore, it provides a relatively objective measure to analyze the main themes of massive corpora of abstracts more comprehensively and systematically.
Contemporary science is described as a complex network whose vertices are the elements of the system, and whose edges represent the interactions between them. [27]ariety, balance, and similarity are three components of interdisciplinarity. [28] In this project, network analysis is used to measure the "interdisciplinarity" of publications based on their WoS subject categories and scale the integration of collaborative networks at the institutional level.
The CNCI metric is a widely used indicator to benchmark an institution's research performance. [29]CNCI was utilized to calculate the institution performance in the AIE field in order to divide the actual count of citing items by the expected citation rate for documents with the same document type, year of publication, and subject.

Growth of the scientific knowledge and cognitive extents in the AIE field
Pooling all the WoS papers on AIE on a year-by-year basis from 2001 to 2021 revealed that the number of publications and citations increased exponentially (Figure 1).The number of papers increased rapidly from 1 to 2043, and exponential growth of publications was seen since 2001 (Figure 1B).The first report published in 2001 on AIE was cited by six publications in 2002, and the number of its cumulative citing times was 5525 till the end of 2021.The number of cumulative cited papers increased more substantially from 5525 to 410557 (Figure 1A).
The growing volume of scientific research and its citing behavior suggest that research on AIE is emerging, and its impact has been growing rapidly.However, the rapid increase in publications and their citations does not reflect the expansion of the cognitive boundaries of science.Cognitive extent is a more precise measure of the development of scientific fields than publication growth. [24]In addition to productivity-based metrics, we measured the cog-nitive growth of phrases rather than words from a dataset in which synonymous keywords and terminologies were merged.Figure 1C illustrates that the cognitive territories have been increasing.This increase was much faster from 2013.The emergent new keywords in 2013 were more related to nanotechnology or biomedicine, such as biocompatible nanoparticle, biogenic-amine, nanoprobe, piezochromic luminescence, polymer sensor and sensitive recognition.It shows that the research on AIE turned to focus more on their practical applications in biomedicine and materials.The subject categories were assigned to individual papers in the WoS.The number of subject categories also increased from one to 33, which reflects that the remarkable increase in publications on AIE did not engender specialization but fostered interdisciplinary research.The exponential growth of institutions (Figure 1D) reflected the rapid diffusion of AIE across time and space.

Institution and country diversity
Are AIE sciences fragmented because of enormous growth in scholarly output?If this were the case, the diffusion of ideas would have been more difficult.However, diversity in the composition of research institutions promotes new concepts from diverse perspectives anchored in different geographical or institutional entities.In addition to the paperlevel measure, institution-and region-level publications were counted to measure the institutional diversity and collaborative features of institutions studying AIE.How many types of institutions and how different from each other are the types of items that are measured.The analysis of the geographical distribution revealed the dynamic patterns of research leadership at the institutional and country levels.Figure 2 illustrates that AIE researchers are distributed across 72 countries, among which China, the United States of America, India, Japan, and South Korea are the top five at the institutional level, while China, India, Japan, the United States of America, and Singapore are the top five at the paper level (Table S1).The general trend of the top 10 productive research institutions in the field of AIE (Figure 3) is rapidly increasing, while there was a fast increase from 2008 and a slight slowdown from 2020 because of the pandemic impact.The Hong Kong University of Science and Technology and the South China University of Technology are the top two productive institutions in China.The National University of Singapore and the Indian Institute of Technology System are the two most productive institutions outside China.
Our analysis indicates a remarkable diversity in the composition of global research institutions.The new concept of AIE drives trends at different institutional levels, and new ideas around AIE are being disseminated and shared by a growing scientific community.

International collaborative network and performance
International collaboration is challenging and costly but often leads to more impactful innovations through concepts and methods that share collaborative learning across geographical distances.To gain an in-depth understanding of the integration of new ideas and methods toward long-distance international collaborations in the field of AIE, a collaboration network was mapped.Considering that researchers in China contribute substantially (84%) to AIE field publications, national collaborations in China are removed to specify the intensity of international collaboration.In the collaboration network, the direction of an edge indicates the collaboration between two institutions, whereas the weight on the edge represents their collaborative intensity.Figure 4 presents the institution-level weights, while Figure S1 shows their evolution in four periods.The Hong Kong University of Science & Technology, UDICE-French Research Universities, National University of Singapore, Indian Institute of Technology System, and State University System of Florida afforded many more collaboration partners.The Hong Kong University of Science and Technology and the National University of Singapore provided the strongest international collaboration (Figure 4 and Figure S1).When all the regional collaborations are removed, Durham University, Royal Institute of Technology appeared in the top 30 institutions which provided strongest international collaboration (Figure S2).
To further investigate the performance of institutions, we applied the category normalized citation impact (CNCI) metric to analyze the field of AIE.As a standard indicator for institutional comparisons, the advantage of the CNCI is that it does not assign potentially incorrect credit to institutions. [29]gure 5 shows that the National University of Singapore retained its highest CNCI value.Our findings are consistent with the fact that most research leadership flow occurs within a small radius. [30]This demonstrates that international collaboration deepens knowledge and leadership flow through AIE concept spanning.This new concept of AIE has motivated broad international research.

Topic modeling AIE fields and the emergent concepts
A dataset of keywords and abstracts from all 10671 publications in the AIE field was collected for the LDA analysis.The topic modeling algorithm in MALLET (LDA) was applied to the dataset.The topics were identified from the top words, and the corresponding papers were checked and revised as emerging topics.Table S2 shows the number of topics and  the parameters.Table S3 presents the identified topics and the most representative articles for each topic.As a result of topic modeling (Tables 1 and 2), AIE significantly accelerates topics spanning from single AIE molecules (topic # A1), the mechanism causing the AIE effect (topic # A2) to designing and modulating highly efficient luminogens (topics # B3, B7, B9, B10, C6, C8, C10, D4, D6, D11, D13, D14, and D17), aggregation formation, AIE processes   The distribution of topics across four periods indicates that the proportion of physical processes and theoretical research was larger than that of AIE material design and its applications.A clear mechanistic explanation of the AIE effects provides accurate predictions of the common properties of a series of luminogens with similar structural features.Guided by mechanistic understanding, researchers have designed new AIE systems that offer new materials for further studies.These new materials then helped discover new photophysical phenomena and collect additional mechanistic information.This is why the AIE field has increased drastically regarding publications, citations, cognitive extent, and institutional diversity, as indicated in the above data analysis outcome.
The evolution of topics reveals that it is neither theorydriven nor technology-driven, but that the theory-practice integration process drives the progress of the AIE field.In 2001, Tang and his team observed a diametrically opposed phenomenon.After that, they studied AIE in three strategies at the same time.These are developing new AIEgens, studying the mechanisms and finding their practical applications, which integrated theoretical and practical research.As shown in Table 1, topic # A1 and # A2 are about designing new AIEgens and studing the mechanism causing the AIE effect theoretically.Topics on practical applications (topics # B6, B12) were immediately appeared in the second period (2006-2010).It is more obviously that these three strategies are integrated deeply in 2011-2021, which are shown in Table 2.For instance, the theoretical study on the mechanisms revealed that restriction of intramolecular motions (RIM) is responsible for the AIE processes.It provied a fundamental theory for scientists to find alternative mechanism for restricting the rotation of the phenyl rings when they designed the new metal-organic frameworks (MOFs) and studies tight packing of the tetraphenylethylene chromophores in 2011. [31]Meanwhile, these finding suggested the potential applications of the new MOFs and inspired the following studies to find a large array of practical applications.
Consequently, the concept of AIE has been proposed by chemists, but the research scale in this field is typically F I G U R E 5 Top 10 institutions with high category normalized citation impact (CNCI) values in the field of aggregation-induced emission (AIE).

F I G U R E 6 Emergent concepts distribution in the field of aggregation-induced emission (AIE).
related to mesoscale materials, such as nanoparticles (topics # B5 and C7) and supramolecules (topic # C6).Intramolecular rotation and intermolecular motion of materials in the aggregate state frequently appear as topics (topics # A2, B2, C2, and D5).This helps scientists widen their search avenues toward the microscale and macroscale.Researchers have designed and synthesized new AIE molecules for use as efficient light emitters at the microscale or molecular level.At the macroscale or real-world levels, scientists can explore the enormous application potential of AIE luminogens.Thus, the effect of AIE bridges the huge gap between microscale molecules and macroscale materials, offering a new platform for interdisciplinary studies and changing the traditional metaphysical assumptions centered on a specific molecular level.The concept of AIE broadens the scientists' exploration of an enormously complex world with many structures.For instance, researchers explored a supramolecular approach toward mesogens showing AIE effect and investigated their mesomorphic behavior. [32]o further explore in what sense the paradigm shifts with a change of theory in revolution, emergent concepts were measured according to the frequency of new keywords that emerged in each publication year.Figure 6 indicates that the aggregates/aggregate materials/related mechanisms that dominate the photo-physical properties of aggregate materials constitute the core theories and investigating entities in the AIE field.Core theories provide a means for manipulating a wide variety of engineering and biological processes.From the molecular level to the aggregate level, different aspects of science, such as crystallography, nanotechnology, supramolecular science, and engineering science have been positively involved.The Kuhnian revolution occurred in the AIE field because scientists are working in the world, focus-ing on the properties of aggregate materials, rather than the traditional study of the relationship between molecular structure and properties.Emerging topical fields borrow heavily to expand, such as "activated delayed fluorescence" and "tetraphenylethylene".The centrality of the molecular "tetraphenylethylene" in the AIE field is not merely originating from its structure and properties, but also from the tetraphenylethylene skeleton with the most versatile AIEgen designed for various desired functions.The emergent concept of "amyloid aggregation" in 2021 demonstrates that the concept of AIE brings new perspectives for investigating biological processes at the mesoscale.The continuous expansion of the emerging words in different scales is observed from "intramolecular charge-transfer" to "nanoparticle" and "living cell", which are corresponding to microscale, mesoscale, and macroscale.

Measurement of interdisciplinarity of works and knowledge flow in the AIE field
Given the concept of AIE, which gathers scientists from multiple sciences to work in a new world, the interdisciplinarity of publications is quantified to analyze how works at the mesoscale stimulate collaboration across disciplines at the microscale and macroscale.The WoS lists 171 subfields in science and engineering, 54 in social science, and 27 in the arts and humanities.Subject categories were assigned to individual papers.This classification system has been widely used to measure interdisciplinarity.The number of scientific branches in the cited works on AIE has increased remarkably.Citing papers are mainly distributed in the subject categories of "Chemistry, multidisciplinary", "Material science, multidisciplinary", "Chemistry, physical", "Nanoscience & nanotechnology", and other branches (Figure 7).This reveals the large extent to which AIE draws on knowledge from distinct fields.
Figure 8 disaggregates the dataset of citing publications in WoS on a field-by-field basis, revealing a marked similarity in increasing trends across the branches of science regarding the main findings in the accumulated trends of publication, topic modeling, and emergent concept analysis.This suggests that the AIE field sustains a very high level of interdisciplinarity and coherence.Previous studies indicate that boundary-spanning agendas that bridge disciplines tend to form their discipline-like fields and suggest that boundary-spanning research does not necessarily involve new institutional forms. [33]Our findings in the AIE field support these previous findings.This conclusion is based on the following three observations.First, the attribution of interdisciplinarity of earlier research (2001-2005) on AIE stimulated many more research fields in the following three periods.Second, the connectivity between sub-disciplines increased significantly from year to year.Third, the citation network has an increased connection in "chemistry, physical" and "biochemical research methods", which indicates AIE has emerged as a critical tool not only for advancing our understanding of biological processes but also for facilitating the development of diagnosis and therapy. [18]ow can AIE sustain a high level of interdisciplinarity? Recent quantitative research indicates that fields of science have become more integrated over time, and this observation is attributed to the growth of cross-field communication throughout the entire period, as well as the growing importance of high-impact papers to bridge networks in the same F I G U R E 8 Continued year. [8]To examine the clustering of scholarly communication, we studied the strong links between the top 20 most cited papers on AIE and their citing behaviors over time.As illustrated in Figure 9, limited but high-impact papers form a core network.Aggregating the citation network that expresses how similar they are and the iteration of the key nodes forming pairs in the citation network allows us to study structures distinct from the AIE field.Amazingly, highimpact papers were published in different periods, and 70% of them were published recently, creating unanticipated new directions for AIE research.The finding on the above citation connectivity is consistent with previous measurements of disruptive work in which future papers citing a focal paper ignore its acknowledged forebears. [4]This study provides further evidence that top papers promote the growth of cross-field communication because they integrate theory and application in novel directions.Taking the top paper (https://doi.org/10.1039/C4CS00444B)published in 2015 [34] as an example, this study discusses the operation mechanism, application of AIE fluorogens, and specific light-up bioprobes with advanced functionalities in bioimaging and detection, which have attracted significant attention in the fields of instrumentation, biology, engineering, chemistry, materials science, applied physics, and beyond.For instance, researchers synthesed a four-armed amphiphilic copolymer, in which the tetraphenylethene derivative acted as an AIE flu-orescent probe for live cell imaging and another part was an anticancer drug. [35]

CONCLUSION
Our data analysis revealed that there is now a selfdenominated and fast-growing community of the AIE field-or epistemic community-with a relative cohesion of principles, albeit with some differences between chemistry, physics, and material sciences.It is an interdisciplinary science, which is different from a multidisciplinary or transdisciplinary approach treating a problem by relying on radical reorganization or even dismantling of the current disciplinary motif.Focusing on mesoscale entities, scientists from many different branches of science are involved in theoretical research on mechanisms, as well as developing different AIE systems for applications.The metaphysics and new theoretical assumptions in the AIE concept are that substances in the aggregate state possess unique properties that cannot be reduced to the molecular level.This unusual assumption promotes disruptive work in the AIE field.The observation of the emergence and evolution of the AIE field corresponds with a new theory in the philosophy of scientific practice, which advocates that the world has many structures, but no fundamental or all-encompassing structures. [6]This differs from the ideal of unifying science and the argument that disciplines, including chemistry, are reducible to physics or methodological reductionism; the properties of entities at the mesoscale or macroscale can be reduced to explanations on a microscale.The data analysis in this project provides details showing how non-reductionist concepts based on new scientific discoveries cross traditional disciplinary boundaries and aggregate interdisciplinary research.Initially, the AIE field was interdisciplinary, connecting the microscale and macroscale worlds.The continuing interaction between the mechanisms and explanations of the properties of entities in the aggregate state and their applications has attracted more disciplines.Scientists are working in a new scientific world after disruptive work with new core principles, which conceptually and practically links interdisciplinary research.
The data analysis of the disruptive work on AIE inspired us to reconsider the problems in contemporary interdisciplinary research.Besides specialization in disciplines, the traditional boundaries of explainable scales restricted to specific disciplines may also be crossed.Given the variability in scientists' philosophical views challenging interdisciplinary integration across scientific branches, [36] success in the AIE field implies that scientists may be motivated to embrace nonreductionist ideas at different research scales, leading to a more permeable field boundary.

A C K N O W L E D G M E N T S
This work was funded by National Natural Science Foundation of China (grant numbers: 22125103 and 21971065), Science and Technology Commission of Shanghai Municipality (grant numbers: 22JC1401000 and 20XD1421500), the National Social Science Fund of China (grant number: 19BZX041), and the Fundamental Research Funds for the Central Universities (grant number: 2022ECNU-XWK-ZX07).

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare no conflict of interests.

D ATA AVA I L A B I L I T Y S TAT E M E N T
The data used and/or analyzed during the current study are available at https://osf.io/5b6sy/?view_only=, a89b556e66a8426291358c270db2bcde.

F I G U R E 1
Distribution of publications, citations, and cognitive extents in the aggregation-induced emission (AIE) field.(A) Publications and cumulative citations.(B) Trends in the number of publications which increases by exponential growth (R 2 = 0.991).(C) Research fields and keywords distributions.(D) Trends in the number of research institutions which increases by exponential growth (R 2 = 0.9935).

F
I G U R E 1 Continued F I G U R E 2 Geo-distributed network of regions (abbreviation) in the field of aggregation-induced emission (AIE).The size of the node represents the number of publications in this region.Regions with the top 15 institutions number are marked with their institutions' number/publications' number.The bottom left picture shows the scale-up regions.

F I G U R E 3
Top 10 most productive research institutions in the field of aggregation-induced emission (AIE).TA B L E 1 Research topics during 2001-2010.

F I G U R E 4
International collaboration network of the aggregation-induced emission (AIE) field.The size of the nodes represents the number of collaboration partners, and the link weight represents the number of collaborative publications.(A) Institutional-level collaboration network without national collaboration in China.(B) Top 30 research institutions with high collaborative intensity, the size of the words and the nodes both represents the number of collaboration partners.

F I G U R E 7
Top 20 citing papers on aggregation-induced emission (AIE) in WoS research fields and their distribution pattern.

F 9
Historiograph on 20 core publications (top 20 cited publications) and their interactions.The corresponding author, published year, and DOI (Digital Object Identifier) name are listed in the node.

Time periods Number Topic identified by experts Top 10 words 2001
-2005 A1 Silole exhibiting high emission in aggregate state silole exhibiting high emission in aggregate state A2 Intramolecular rotation restriction in soluble polymers exhibiting AIE solution, induce, high, intramolecular, layer, yield, compound, fluorescence, restrict, temperature Research topics during 2011-2021.