Key Traits of Leading Sustainable Firms in the Semiconductor Industry

: Few studies have investigated sustainable management in the semiconductor industry. Consequently, this study analyzed the characteristics of companies excelling in sustainable management in the semiconductor industry using chief executive officer messages. It compared high-and low-performing groups to identify leading sustainable firms. Centrality analysis was conducted to extract keywords, which were mapped to the sustainability criteria to conduct network analysis. The results showed that the high-performing group emphasized sustainable development across the semiconductor industry ecosystem, while the low-performing group focused on internal sustainability aspects. This underscores the need for effective sustainable development in the semiconductor industry that extends beyond individual company efforts and embraces industry solidarity. Thus, this study presents a methodology that can be applied to similar studies in industries beyond semiconductors.


Introduction
The semiconductor industry stands as a cornerstone of modern technological advancement, underpinning innovations across a vast array of sectors [1].In recent years, however, the industry has come under increasing scrutiny due to its environmental and social impacts, prompting a paradigm shift toward sustainability [2].This study aims to illuminate the distinguishing traits of high-performing companies within the semiconductor industry, underscoring their significance amidst the evolving global sustainability imperatives.
Within the broader context of corporate sustainability, the semiconductor sector represents a unique nexus where cutting-edge technology intersects with pressing environmental and social concerns [2].While the imperative for sustainable practices within the industry is widely acknowledged, the precise determinants of sustainability performance remain subject to debate [3].Existing literature offers valuable insights into various aspects of sustainability management within the semiconductor industry, highlighting the multifaceted nature of this challenging task [4].
Notably, diverging hypotheses exist regarding the key drivers of sustainability performance among semiconductor companies [4].Some scholars argue for the primacy of technological innovation and operational efficiency in driving sustainability outcomes [5], whereas others emphasize the importance of supply chain transparency and stakeholder engagement [6].Additionally, the role of regulatory frameworks and industry standards in shaping sustainability practices remains a subject of contention, with implications for organizational strategies and performance [7].
Over the past three decades, the measurement of sustainability has emerged as an exceptionally captivating subject among researchers [8], being widely acknowledged as a crucial issue with a significant impact on the formulation and implementation of corporate strategies [9].Sustainable development is defined as "development that meets the needs of the present without compromising the ability of future generations to meet their own needs" [10], thus necessitating a more refined understanding across various domains [11].Business sustainability, encapsulated by the concept of the triple bottom line [12], pertains to the preservation of future human and natural resources while concurrently addressing the current needs of businesses and stakeholders [13].This entails a holistic consideration of environmental, social, and economic parameters [14,15].
In recent years, the discourse surrounding sustainability performance has garnered significant scholarly and practical attention, particularly driven by the imperatives set forth by global initiatives such as the Sustainable Development Goals (SDGs) [16][17][18].The SDGs serve as a comprehensive framework encompassing a wide array of environmental, social, and economic sustainability objectives aimed at guiding global efforts toward a more sustainable future [17].Within this context, private enterprises have increasingly recognized the importance of integrating sustainability considerations into their operational strategies [18].Consequently, various methodologies have emerged to assess and evaluate sustainability performance, catering to the diverse needs and contexts of different industries and sectors [16,18].Among these methodologies, the utilization of scoring systems to measure the ESG criteria has gained prominence [19,20].Notably, Sustainalytics has emerged as a leading provider of such ESG ratings, offering comprehensive assessments of companies' sustainability performance based on a multitude of factors [20,21].Through its rigorous evaluation process, Sustainalytics provides valuable insights into the sustainability practices and initiatives of companies, facilitating informed decision-making by stakeholders, ranging from investors to consumers [20].

Trends in Sustainability within the Semiconductor Industry
The semiconductor industry, characterized by its pivotal role in driving technological innovation across various sectors, has come under increasing scrutiny for its environmental and social impacts [22].Scholars and practitioners alike have recognized the need for sustainable practices within this industry to mitigate these impacts and ensure long-term viability [23,24].Consequently, research in this domain has explored various aspects of sustainability, ranging from specific manufacturing processes and material treatments [25,26] to broader industry-wide initiatives aimed at reducing the environmental footprint [27] and enhancing social responsibility [15,28].
There has been considerable scholarly discourse regarding the criteria and frameworks for achieving sustainability, including the incorporation of the SDGs [29,30].In the semiconductor industry, endeavors have been undertaken to establish criteria for achieving sustainability [31].One research has yielded 13 criteria across the environmental, economic, social, and innovation dimensions as part of these sustainability standards [11].These serve as a set of comprehensive guidelines for assessing and benchmarking companies' sustainability performance [11].Moreover, these criteria encompass a wide range of factors, including resource efficiency, emissions reduction, supply chain management, and social responsibility [11].In the current study, we utilize these 13 sustainability criteria within the semiconductor industry to analyze differences between the high-performing group (HPG) and low-performing group (LPG) in the field of sustainable management.

Trends in Research on Sustainability Reporting
With the increasing emphasis on sustainability performance [32], companies are now required to transparently communicate their sustainability efforts and outcomes to stakeholders [33,34].In other words, as sustainability reporting is becoming mandatory [35], the importance of sustainability reports is growing simultaneously [36].Sustainability reports serve as comprehensive documents that detail companies' ESG practices, initiatives, and performance metrics [37].These reports are crucial in enhancing accountability, fostering trust, and guiding decision-making among stakeholders, including investors, consumers, employees, and regulatory bodies [38,39].
Of particular significance within sustainability reports is the inclusion of CEO messages, which often serve as the introductory or concluding sections of these documents [40,41].CEO messages provide a platform for company executives to articulate their vision, values, and commitment to sustainability, thereby setting the tone for the entire report [42].Moreover, CEO messages offer insights into the companies' strategic priorities, challenges, and opportunities related to sustainability management, thus shedding light on their long-term sustainability strategies and aspirations [43,44].

Trends in Sustainability Reporting within the Semiconductor Industry
Recent research has explored sustainability reporting in the semiconductor industry.For instance, studies have evaluated the sustainability of Taiwan's semiconductor industry [45] and examined supply chain transparency [46].Additionally, studies have investigated factors influencing the adoption of sustainability reporting adoption in China's semiconductor industry [4], explored the relationship between ESG factors and financial performance in Malaysian semiconductor firms [47] and ESG management of semiconductor equipment production [48], and examined greenhouse gas(GHG) emissions in semiconductor companies [49].
Nonetheless, existing research has often targeted specific countries or focused on limited aspects of sustainability, thus necessitating broader studies encompassing the entire semiconductor industry.This study fills this gap by examining semiconductor companies worldwide, without restricting its scope to a specific country, and investigating various aspects of sustainability, including ESG considerations, across all sectors beyond the semiconductor industry.Furthermore, this study compares and analyzes the characteristics of leading and lagging companies in sustainability within the semiconductor industry, aiming to address the following research questions: (1) What are the distinguishing features of leading sustainability companies in the semiconductor industry?(2) What efforts should companies, governments, and relevant institutions focus on to foster sustainable development in the industry?

Analysis Subject and Research Procedure
With the increasing emphasis on sustainability performance [32], companies are now required to transparently communicate their sustainability efforts and outcomes to stakeholders.The analysis in this study focuses on 92 global semiconductor companies identified by Sustainalytics as of 31 December 2023, that meet the following criteria: companies that (1) published sustainability reports during 2023, (2) include CEO messages in their sustainability reports, and (3) are rated by Sustainalytics for ESG risk, excluding those rated as Risk Medium.This exclusion aims to starkly contrast companies with excellent and inadequate sustainability scores.Sustainalytics' ESG rating evaluates ESG risks, categorizing companies based on their sustainability performance into four tiers: Risk Low, Risk Medium, Risk High, and Risk Severe.Consequently, the highest-rated companies, exhibiting exemplary sustainable management practices, are classified under Risk Low, followed by Risk Medium, Risk High, and Risk Severe, in descending order of sustainability performance [50].Based on these Sustainalytics' ESG rating criteria, the 92 companies were identified and categorized into the following HPG and LPG for comparison: HPG: 55 Risk Low firms and LPG: 37 Risk High and Severe firms.Although the number of companies in each group varied, the two groups were independent samples, and both groups exceeded a sample size of 20, allowing for meaningful comparative analysis of the characteristics of the two groups [51].
Moreover, this study conducted network analysis using keywords from CEO messages and criteria mapping.The study procedure involved six stages (Figure 1).

Data Collection
The analysis cohort comprises 92 global semiconductor companies identified by Sustainalytics within the semiconductor industry as of 31 December 2023.This selection involved excluding 62 companies that had a Risk Medium ESG rating or lacked CEO messages in their sustainability reports.Sustainalytics' ESG rating system evaluates companies on their sustainability practices, categorizing them as Risk Low for exemplary performance, Risk Medium for moderate performance, Risk High for inadequate performance, and Risk Severe for severely deficient performance [50].In this study, 55 companies rated as Risk Low were designated as HPG, while 37 companies rated as Risk High and Risk Severe were categorized as LPG for comparative analysis.All 92 companies underwent Sustainalytics' ESG risk evaluation in 2023 and had a documented history of publishing sustainability reports that year.Tables A1 and A2 in Appendix A present the performance data for firms assessed as high and low performers according to Sustainalytics' ESG Risk evaluation as of 31 December 2023.

Data Collection
The analysis cohort comprises 92 global semiconductor companies identified by Sustainalytics within the semiconductor industry as of 31 December 2023.This selection involved excluding 62 companies that had a Risk Medium ESG rating or lacked CEO messages in their sustainability reports.Sustainalytics' ESG rating system evaluates companies on their sustainability practices, categorizing them as Risk Low for exemplary performance, Risk Medium for moderate performance, Risk High for inadequate performance, and Risk Severe for severely deficient performance [50].In this study, 55 companies rated as Risk Low were designated as HPG, while 37 companies rated as Risk High and Risk Severe were categorized as LPG for comparative analysis.All 92 companies underwent Sustainalytics' ESG risk evaluation in 2023 and had a documented history of publishing sustainability reports that year.Tables A1 and A2 in Appendix A present the performance data for firms assessed as high and low performers according to Sustainalytics' ESG Risk evaluation as of December 31, 2023.
To collect data, the researchers visited the websites of all companies in both groups and downloaded the PDF or HTML versions of their 2023 sustainability reports.Before collecting the CEO messages, it was assumed that the annual reports primarily focused on financial aspects, while CEO messages in sustainability reports predominantly highlighted non-financial matters, including the company's sustainability philosophy, vision, and performance [40- 44,52,53].CEO messages within sustainability reports were then excerpted, grouped by category, and consolidated into two TXT files-one for each groupto form the ultimate dataset.

Data Preprocessing
Most sustainability reports are published in the native language of the country where the company's headquarters are located or in English.However, among the 92 companies studied, 4 Chinese companies published their sustainability reports only in Chinese, and 1 South Korean company published theirs only in Korean.CEO messages from these five To collect data, the researchers visited the websites of all companies in both groups and downloaded the PDF or HTML versions of their 2023 sustainability reports.Before collecting the CEO messages, it was assumed that the annual reports primarily focused on financial aspects, while CEO messages in sustainability reports predominantly highlighted non-financial matters, including the company's sustainability philosophy, vision, and performance [40- 44,52,53].CEO messages within sustainability reports were then excerpted, grouped by category, and consolidated into two TXT files-one for each group-to form the ultimate dataset.

Data Preprocessing
Most sustainability reports are published in the native language of the country where the company's headquarters are located or in English.However, among the 92 companies studied, 4 Chinese companies published their sustainability reports only in Chinese, and 1 South Korean company published theirs only in Korean.CEO messages from these five companies' sustainability reports were translated into English using Google Translate, a method previously validated in similar research [54].The data collected in TXT format represented unstructured textual data and required preprocessing before analysis [55].For text preprocessing, we utilized Python programming, a well-established method in prior research [56][57][58], with Google Colab's Notebook environment.We standardized the text by converting all characters to lowercase and removing irrelevant numbers, symbols, articles, prepositions, conjunctions, adverbs, and words lacking analytical significance.Additionally, plural nouns were singularized to ensure consistency.These preprocessing steps laid the foundation for meaningful results in the subsequent first network analysis.

First Network Analysis
Unlike traditional text analysis methods that analyze specific concepts or words [54] in a text statistically, language network analysis enables the understanding of relationships between specific concepts in the text [59].Represented as a network, this information enables quantitative analysis; it focuses on the relationships between words rather than solely on word frequency, thereby delving into the implicit context of language [59,60].
In network analysis, words are depicted as nodes, and the connections between them are depicted as links.This structure mirrors that of social network analysis techniques [61].Language network analysis, an expanded content analysis technique, operates with networks where nodes symbolize actors and links denote their connections, akin to standard network analysis [59,62].In this method, crucial terms or phrases within the text are depicted as nodes and their associations as links, and directional data between the words are disregarded [63].Centrality analysis, a quantitative approach, is commonly utilized to uncover pivotal terms within the network.Through the application of diverse centrality analysis methods, terms that prominently feature in the text can be pinpointed, facilitating a deeper understanding of the core messages conveyed.The notion of centrality within a network is fundamental in gauging the significance of specific nodes' central roles within the network structure [64].The numerical values computed for each node in the network represent their relative standings.
In this study, Python was employed within Google Colab to conduct centrality analysis.We investigated the eigenvector, betweenness, and closeness centralities and extracted keywords based on the quantitative values of each centrality measure, from the 1st to the 30th positions [54].Additionally, to mitigate bias toward straightforward connections between nodes, we abstained from utilizing centrality measures reliant on edge centrality values [63].
By leveraging the results of the eigenvector, betweenness, and closeness centralities from the initial network analysis of keywords in each group, we scrutinized the sustainability attributes of HPG and LPG.The primary objective of the initial network analysis was to identify and compare crucial keywords extracted from the CEO messages of each group.

Criteria Mapping
Based on the results of the eigenvector, betweenness, and closeness centrality analysis, we selected keywords within the top 30 in each centrality measure, including duplicates [54].To prioritize these keywords, we assigned weights based on the presence of keywords in each of the three centrality measures.If a keyword appeared in the top 30 for all three measures, it received a weight of 3. If it appeared in the top 30 for two measures, it received a weight of 2, and if it appeared in the top 30 for only one measure, it received a weight of 1 [54].
Having compiled a roster of keyword contenders along with their respective weights, we identified keywords carrying a weight of 3. Subsequently, we mapped and categorized these selected keywords against sustainability criteria to scrutinize their distinctive attributes.The sustainability criteria were selected by referencing recent literature in the semiconductor industry, with a particular focus on sustainable development aspects.The 13 sustainability criteria borrowed from this source include (1) clean energy use, (2) recycling/renewable capacity, (3) green resource integration, (4) pollution-discharge treatment, (5) firm size, (6) financial strength, (7) material cost/selling price, (8) partner complementarity, (9) corporate brand image, (10) customer relationship management (CRM) capability, (11) core technical patent, (12) product life cycle, and (13) research and development (R&D) capability [3].
The relationship between words and criteria was established by searching for sentences in the CEO messages where the keywords identified through centrality analysis appeared.This mapping considered the lexical meaning of the words, the context within the entire paragraph, and the context within the specific sentence [65].Each word was mapped to three criteria, and if it was associated with more than four criteria, we prioritized and mapped it to the top three criteria, ensuring the relevance of the mapping results [54].The process of mapping keywords to sustainability criteria was confirmed through interviews with 15 experts with more than five years of experience in the semiconductor industry, thus enhancing the objectivity of the mapping results.

Second Network Analysis
By leveraging the mapping results, we treated the mapped criteria as individual nodes and established links between them to analyze the connectivity in the secondary network analysis.
For our analysis, we utilized Gephi 10.1, a popular open visualization software widely used for network analysis.We constructed a one-mode matrix for each group with sustainability criteria serving as nodes and visualized the matrix using Gephi 10.1 by connecting the nodes with edges.
Subsequently, using the weight-3 keywords and mapping outcomes from the criteria mapping, we compared the sustainability attributes of HPG and LPG.The second network analysis aimed to identify and contrast crucial sustainable management factors derived from the CEO messages of each group.

Centrality Analysis
The HPG dataset comprised 55 firms, 1644 sentences, and 37,475 words.Table 1 lists the top 30 words along with their centrality scores for the three centrality analyses for HPG.The most significant words-"sustainability", "company", "global", "product", "technology", and "employee"-were highlighted.These words were deemed crucial as they ranked within the top 10 across all centrality metrics.The LPG dataset consisted of 37 firms, 978 sentences, and 24,161 words.Table 2 lists the top 30 words along with their centrality scores for the three centrality analyses conducted specifically for the LPG.The most crucial words include "development", "company", "energy", "employee", "business", "management", "esg", and "product".These words were considered vital as they ranked within the top 10 across all centrality metrics.

Weighting Keywords
Based on the centrality analyses, we assigned weights to the words according to the frequency of their appearance in the top 30 among the three centrality results [54].For the HPG, 27 words were assigned a weight of 3, 2 words were assigned a weight of 2, and 5 words were assigned a weight of 1. Table 3 lists these words, grouped by their respective weights.In LPG, 26 words were assigned a weight of 3, 4 words were assigned a weight of 2, and 4 words were assigned a weight of 1. Table 4 presents words grouped by their respective weights.

Keyword Mapping with the Sustainability Criteria
In the HPG study, mappings were conducted with the sustainability criteria [4] for the semiconductor industry by focusing on words with a weight of 3. "Green resource integration" emerged as the most frequent criterion with 20 occurrences, followed by "partner complementarity" with 14 occurrences and "R&D capability" trailing with 10 occurrences (Table 5).The remaining criteria had less than 10 occurrences, and "financial strength", "material cost/selling price", and "core technical patent" were not mapped even once.
In the LPG study, mappings were conducted with the sustainability criteria [4] for the semiconductor industry by focusing on words with a weight of 3. "Corporate brand image" emerged as the most frequent criterion with 24 occurrences, followed by "pollutiondischarge treatment" with 16 occurrences and "recycling/renewable capacity" trailing with 12 occurrences (Table 6).The remaining criteria had less than 10 occurrences, and "financial strength" and "core technical patent" were not mapped even once.

Criteria-Based Network Analysis and Visualization
In the HPG study, a one-mode matrix was constructed based on the mapped criteria (Table 7).The four most influential nodes were identified as "green resource integration," "partner complementarity", "R&D capability", and "recycling/renewable capacity".Upon analyzing the relationships between these nodes, the following insights were gleaned: (1) "Green resource integration" exhibited strong connections with "partner complementarity", "recycling/renewable capacity", and "CRM capability".( 2) "Partner complementarity" demonstrated a strong connection with "green resource integration".(3) Although "R&D capability" did not exhibit particularly strong connections with any specific node, it connected diversely with all nodes.(4) "Recycling/renewable capacity" showed strong connections with "corporate brand image", "green resource integration", and "pollution-discharge treatment".Figure 2 visualizes these findings.
Similarly, in the LPG study, a one-mode matrix was also constructed based on the mapped criteria using the same method (Table 8).The five most influential nodes were identified as "corporate brand image", "pollution-discharge treatment", "recycling/renewable capacity", "CRM capability", and "clean energy use".Upon analyzing the relationships between these nodes, the following insights emerged: (1) "Corporate brand image" exhibited strong connections with "pollution-discharge treatment", "recycling/renewable capacity", "clean energy use", and "CRM capability".( 2) "Pollution-discharge treatment" showed strong connections with "corporate brand image" and "recycling/renewable capacity".

Discussion
Previous sustainability studies within the semiconductor industry have pred nantly focused on case studies of specific companies or countries rather than sustaina performance or evaluation [45][46][47][48].Furthermore, research in this field has often been ited to specific processes such as chemical mechanical polishing [25] or specific issue GHG emissions [23].By contrast, this study conducted a comprehensive examinati the overall characteristics of leading sustainable companies within the semiconduct dustry.This study undertook a network analysis of CEO messages, categorizing them HPG and LPG to identify the traits of HPG firms.

First Network Analysis Result
The analysis of keyword centrality revealed no significant difference between and LPG.When analyzing the top 30 keywords for each group based on centralit found a considerable overlap.Specifically, in terms of betweenness centrality, 25

Discussion
Previous sustainability studies within the semiconductor industry have predominantly focused on case studies of specific companies or countries rather than sustainability performance or evaluation [45][46][47][48].Furthermore, research in this field has often been limited to specific processes such as chemical mechanical polishing [25] or specific issues like GHG emissions [23].By contrast, this study conducted a comprehensive examination of the overall characteristics of leading sustainable companies within the semiconductor industry.This study undertook a network analysis of CEO messages, categorizing them into HPG and LPG to identify the traits of HPG firms.

First Network Analysis Result
The analysis of keyword centrality revealed no significant difference between HPG and LPG.When analyzing the top 30 keywords for each group based on centrality, we found a considerable overlap.Specifically, in terms of betweenness centrality, 25 keywords were found to be common between the two groups.Similarly, in terms of closeness and eigenvector centralities, 24 keywords overlapped between the groups.Even when restricting the analysis to the top 30 keywords with a weight of 3 for the eigenvector, betweenness, and closeness centralities, similar results were observed.Among the 27 weight-3 keywords identified in HPG, 23 were also present in the weight-3 keywords of LPG.This can be interpreted as a result of the frequent occurrence and high influence of certain keywords that are inherent to the nature of sustainability reporting documents.

Second Network Analysis Result
In contrast, a significant difference was observed in the mapping of semiconductor sustainability criteria.This disparity stemmed from the contextual meanings of keywords within the sentences in each group.For instance, one HPG sentence-"We have completed the first three-year sustainability audit in 2022, and provided suppliers with guidance to make improvements, which have been completed across our supply chain"-used the term "sustainability" to denote improvement not only within the company but also across the supply chain, aligning with the "green resource integration" criterion.Conversely, the use of "sustainability" in LPG sentences-such as "Over the years, we have always adhered to the concept of sustainability to create a responsible, green and low-carbon corporate image"-emphasized the corporate image regarding the sustainability of the company itself; this corresponded with the "corporate brand image" criterion.Thus, even though the same word was used, the contextual emphasis by CEOs in their messages varied.
In the criteria mapping and secondary network analyses, significant differences were revealed between HPG and LPG.In HPG, the most influential criteria were "green resource integration", "partner complementarity", "R&D capability", and "recycling/renewable capacity"."Green resource integration" pertained to how well a company managed sustainable supply chain management [3], with a predominant focus on ensuring all companies within the industry effectively implemented sustainable management practices."Partner complementarity" highlighted the competitiveness of integrating and enhancing various resources, capabilities, and technologies possessed by all stakeholders in the semiconductor industry [3], underscoring the importance of resources and capabilities of public institutions, universities, customers, suppliers, and competitors."R&D capability" refers to the ability to innovate based on R&D, ensuring product innovation capability and advanced technology [3] and suggesting the continued significance of R&D innovation in sustainable management."Recycling/renewable capacity" indicates the ability to create environmentfriendly products using recycling and renewable technologies [3], reflecting the importance of producing eco-friendly products through environment-friendly methods.
In contrast, the most prominently addressed criteria in LPG were "corporate brand image", "pollution-discharge treatment", "recycling/renewable capacity", "CRM capability", and "clean energy use"."Corporate brand image" denoted the value of corporate social perception [3], prioritizing branding efforts to portray high value to the customers and stakeholders."Pollution-discharge treatment" signified the implementation of environmental protection policies by reducing pollution emissions and improving energy efficiency [3], thus emphasizing the minimization of pollution emissions and environmental protection."Recycling/renewable capacity" as mentioned earlier, referred to the ability to create environment-friendly products using recycling technologies [3], emphasizing the production of eco-friendly products using environment-friendly methods."CRM capability" represents the ability to satisfy customer needs [3], treating customer satisfaction as a key value in sustainable management practices."Clean energy use" indicates the ability to establish a wafer fabrication capable of using clean energy through energy-efficient measurement technologies to reduce power consumption [3], thus highlighting the need to establish eco-friendly fabs and use clean energy.These findings underscore the varying priorities and emphases on sustainability criteria in the HPG and LPG contexts.
In the results of the secondary network analysis, a significant difference was observed between HPG and LPG in the scope of emphasis on sustainable management practices.HPG emphasized the importance of practicing sustainable management not only within their company but also by leveraging all resources and capabilities within the industry, including customers, suppliers, competitors, public institutions, educational institutions, and consumers.In contrast, LPG focused on emphasizing their company's environmental and brand image, as well as customer satisfaction.Specifically, HPG prioritized comprehensive sustainable management practices across the upstream and downstream supply chain management within the industry, while LPG emphasized sustainable practices within their own company and customer satisfaction in downstream operations.HPG may have already reached a stage where they have confidence in their internal sustainable management practices and corporate brand image; therefore, their CEOs emphasized providing a vision for the next step in the development of the industry ecosystem.Contrarily, LPG may not have reached this stage yet; hence, their CEOs emphasized internal environmental initiatives and eco-friendly products, as well as corporate image rejuvenation.
The analysis results have the following implications: (1) LPG should take an interest in the development of sustainable management practices across the semiconductor industry ecosystem, beyond just internal environmental activities, and contribute to the most emphasized criterion in HPG-"partner complementarity".( 2) In LPG, it is important to recognize that sustainable practices should encompass the entire supply chain within the semiconductor industry, including suppliers, thus reflecting the philosophy of "green resource integration."Considering the net-zero scope, LPG should not only focus on Scopes 1 and 2 but also consider Scope 3 simultaneously to demonstrate their willingness and philosophy to develop further.(3) The most influential criterion in LPG was "corporate brand image".Therefore, caution should be exercised in ensuring that sustainable management practices do not solely focus on enhancing corporate brand value, thereby avoiding the pitfalls of ESG washing.

Conclusions
This study advances the academic review of sustainability in the semiconductor industry.While previous research relied on traditional methods such as analyzing eco-friendly practices in specific processes, experimental studies, case studies of individual companies, and interviews with semiconductor experts, this research analyzed CEO messages from sustainability reports formally published by companies.This approach highlights the vision and philosophy of leading companies in the semiconductor industry regarding sustainability, as well as insights into how they implement sustainability initiatives.Moreover, the study connected the sustainability strategies described by CEOs with industry-specific sustainability criteria in the semiconductor sector, providing a clearer understanding of their characteristics.Consequently, it offers insights into the key leadership outstanding companies provide in driving sustainability forward.

Implications
Academically, by innovatively integrating text mining, network analysis, and mapping with sustainability criteria within the semiconductor industry, this study has advanced research on sustainability within this industry.By directly referencing sustainability criteria within the semiconductor industry, the research methodology used in this study provides a novel approach for effectively conducting text-based network analysis in specific industries.It therefore offers valuable scholarly insights that can serve as a reference for future sustainability research in other industries.
Practically, the study has yielded significant insights.Leading sustainability firms prioritize their sustainability practices and the entire industry ecosystem.They emphasize sustainable development for all participants within the supply chain reflecting net zero scope 1, 2, and 3 in their sustainable strategy and advocate for collaboration with government agencies, public institutions, academic and educational organizations, and sustainability groups beyond the supply chain.In contrast, sustainability laggards emphasize their sustainable brand image and eco-friendly energy usage.These findings suggest that sustainability followers can contribute to the overall sustainable development of the industry by benchmarking the integrated and collaborative sustainability practices emphasized by leading firms.Moreover, they should integrate these practices into their formulation of sustainable management strategies.By so doing, not only downstream firms but also all stakeholders will be able to understand the sustainable management mechanisms of leading companies, thereby contributing to the sustainable development of the semiconductor industry.
Policy implications have also been provided in this context.For companies to effectively practice sustainable management, assistance from governments or policymakers is necessary to facilitate close networking among businesses, and sustainability-related organizations including non-governmental organizations, educational institutions such as universities, local government agencies, and leading sustainability firms.This will foster a sustainability ecosystem, enabling companies in the vast semiconductor industry to move in the right direction.

Limitations and Future Research Directions
This study was conducted based on each company's sustainability report published in 2023 and Sustainalytics' sustainability ratings as evaluated in 2023.Owing to the analysis being conducted on a single-year basis, it is limited in its ability to track changes in sustainability trends.Therefore, future research on sustainability performance should be conducted over a period of several years to identify the direction of trend changes.
Furthermore, as this study only focused on semiconductor companies, it is uncertain whether the characteristics of leading companies in sustainable management, as revealed in the research results, are applicable to other industries.If similar studies are conducted for other industries in the future, it can shed light on whether the characteristics of leading sustainable management companies vary across industries or if they possess common attributes.

Figure 1 .
Figure 1.Research procedure and method.

Figure 1 .
Figure 1.Research procedure and method.
relationship management; R&D: research and development.
relationship management; R&D: research and development.

Figure 2 .Figure 3 .
Figure 2. One-mode matrix visualization for the high-performing group.CRM: customer relationship management; R&D: research and development.Sustainability 2024, 16, x FOR PEER REVIEW 14

Figure 3 .
Figure 3. One-mode matrix visualization for the low-performing group.CRM: customer relationship management; R&D: research and development.

Table 1 .
Results of the high-performing group's centrality analysis.

Table 2 .
Results of the low-performing group's centrality analysis.

Table 3 .
Word weighting for the high-performing group.

Table 4 .
Word weighting for the low-performing group.

Table 5 .
Mapping results for the high-performing group.

Table 6 .
Mapping results for the low-performing group.

Table 7 .
One-mode matrix for the high-performing group.

Table 8 .
One-mode matrix for the low-performing group.