Koomey's law forevermore? A document analysis of the production and circulation of the promise of ‘green 5G’

We conduct an analysis of a UK-focused corpus of documents that represent key sites through which the promise of green 5G is produced, circulated and challenged. By the promise of green 5G we refer to an emerging, overarching, dominant expectation that 5G will produce environmental sustainability benefits of various kinds. We employ an analytical approach informed by the sociology of expectations and the concept of technoscientific promises to identify: the various types of expectations and promises upon which the promise of green 5G is built, the ways that the legitimacy and credibility of this promise are boosted, the negative expectations that challenge it, the different ways in which 5G is positioned in relation to the problem of expected traffic growth and how such problematisations direct attention to different solutions, and, finally, a number of other performative effects of the promise of green 5G. The promise of green 5G directs attention towards technological options that aim to render traffic growth sustainable through energy efficiency improvements and away from alternative approaches. This dominant expectation that traffic growth can be rendered sustainable through technological innovation suggests strong commitments to the environmental discourses of ecological modernisation and Promethean environmentalism within this technological community.


Introduction: 5G and a green connected world?
5G is the latest mobile network generation that continues to be rolled out around the world. By March 2021, 5G was available in every region of the world, 5G networks covered 15 % of the global population, and 5G accounted for 4 % of all mobile connections (GSMA, 2021). Big expectations have circulated about how 5G might change people's lives and the world in various ways for a number of years (see, for example, Deloitte, 2018;IHS Economics and IHS Technology, 2017). The sociology of expectations emerged around the turn of the century and takes expectationsespecially expectations about emerging technologiesas its object of analysis. The closely related notion of a promise has also received scholarly attention from the field of Science and Technology Studies. A particular set of promises and expectations concerning the sustainability implications of 5G are the focus of this study.
We speak of the promise of 'green 5G' to refer to an emerging dominant expectation that 5G will, in various ways, both improve the environmental sustainability of mobile networks themselves and produce sustainability benefits in other sectors of the economy. The promise of green 5G primarily focuses on the energy use and GHG emissions aspects of environmental sustainability, although some attention has been given to other types of environmental impacts such as resource depletion and toxicity (for example, in NGMN, 2021). The promise of green 5G is, as we will demonstrate, the achievement of a series of more specific promises and expectations about the emerging and future performance and capabilities of 5G (especially concerning its energy efficiency and its capacity to produce 'enablement effects' across various areas of economic activity and social life), alongside broader industry environmental commitments and targets.
The case of the promise of green 5G is significant because, whilst 5G has been widely established to be far more energy efficiency than 4G, anticipated dramatic increases in mobile data trafficat least in part driven by 5G -could potentially counteract efficiency improvements resulting in an overall increase in network energy consumption. This of course occurs amidst the backdrop of increasing urgency in efforts to mitigate the worst effects of climate change, which include commitments on the part of many telcos to adhere to various climate targets. The realisation of the promise of green 5G is therefore a matter of great significance, as are the ways that the promise of green 5G -like all successful promises and expectationsdirects research and innovation efforts towards particular technologies and approaches and away from others. This case is also interesting from the perspective of the sociology of expectations because, notwithstanding some exceptions (e.g. Kriechbaum et al., 2021), most work that has employed this conceptual approach has focused on cases where technologies are in the early stages of their development. The present study is different in this regard because 5G's commercial roll out began during the time period covered by our research.
The structure of the paper is as follows. The following section reviews relevant literature and introduces the concepts of technoscientific promises and technological expectations. We then present our approaches to data collection and analysis in Section 3. This is followed by the results of our analysis in Section 4. Our results include the types of promises and positive expectations expressed within our corpus, the ways in which the legitimacy and credibility of the promise of green 5G are boosted, the negative expectations expressed within our corpus, different problematisation processes and their effect on directing focus and action, and some of the key performative effects of the promise of green 5G visible within our corpus. In our discussion section we summarise the empirical insights that stem from our analysis and draw out some wider conceptual implications of our work, before offering some concluding remarks in Section 6.

Literature and concepts
In this section we first consider relevant literature, before discussing the two closely related concepts that inform our analysistechnoscientific promises and technological expectations. We define these concepts and their relationship to each other before discussing their present-day material effects, and the actors and activates involved in their production and circulation.

Literature
The concepts of expectations and promises have been widely applied to various empirical areas, perhaps chiefly biotechnology, medicine and low carbon technologies. For instance, in relation to biotechnology and medicine, the concept of expectations has been applied to study gene silencing technology (Sung and Hopkins, 2006) and health technology spinoff companies (Lehoux et al., 2017). Turning to low carbon technologies, the sociology of expectations has inspired work on cases such as bioenergy in the UK (Levidow et al., 2014), hydrogen technologies in the US (Bakker et al., 2012), fuel cell and hydrogen vehicles in Germany (Budde et al., 2012;Budde and Konrad, 2019), electric mobility in the Nordic region (Sovacool, 2019), photovoltaic technology in Germany and Spain (Kriechbaum et al., 2018) or Germany and the US (Hoppmann et al., 2020), offshore wind energy in China (Korsnes, 2016), and renewable energy in Germany (Kriechbaum et al., 2021).
Very little comparable work has so far been conducted in relation to 5G specifically. To the best of our knowledge the concepts of expectations and promises have not yet been brought to bear on the case of 5G, let alone the sustainability implications of 5G. The closest work we have been able to identify comes from Mansell, who applies three themes from the work of Chris Freeman to the case of 5G (amongst other ICT innovations) (Mansell, 2021). One of the themes that Mansell pursues is the notion that guiding principles or common-sense practices can become dominant or institutionalised and so shape expectations about ICT design and use, dominant understandings about what is regarded as beneficial or harmful, and pathways of technological innovation that are and are not discovered and pursued (Mansell, 2021). In relation to 5G, Mansell argues that commonly held views that conflate economic and public values and prominent debates concerning national security and the role of Chinese owned companies work to condition societal expectations to predominantly focus on national security issues whilst largely overlooking broader issues around data protection, privacy and surveillance (Mansell, 2021). Whilst Mansell does not directly draw from the sociology of expectations, she nonetheless provides an account of how dominant guiding principles and common understandings are both shaping the types of expected harms that are and are not focused on in (Western) public debate, and obscuring alternative ways of designing and operating digital technologies geared towards the public good rather than commercial datafication (Mansell, 2021). In other words, dominant cultural norms and understandings shape debates and expectations about technologies (in particular the issues that are focused on and those that are overlooked) and can have material effects on innovation trajectories.
Relatedly, Mansell and Plantin conduct a content analysis of nine British newspapers (a mix of broadsheets and tabloids) and their coverage of 5G between January 2017 and March 2020 (Mansell and Plantin, 2020). Their focus was on which issues the newspapers covered and which received little or no coverage. Their findings include that newspapers were more likely to emphasise negative aspects of 5G; there was a dominant focus on particular applications (e.g. enhanced mobile broadband, self-driving cars and smart devices); risks and benefits were underreported across both broadsheets and tabloids; and reporting focused on security concerns surrounding Huawei's potential role in building the UK's 5G infrastructure but largely overlooked wider concerns about surveillance, privacy and data protection (Mansell and Plantin, 2020).
Some work has been conducted on expectations and promises relating to ICT more broadly. For instance, in a discussion of how particular promises draw on the power of the ideograph of 'technological progress', 1 van Lente illustrates the point through the case study of the development and adoption of High Definition Television (HDTV) in the 1990s ( van Lente, 2000). Van Lente charts how past instances of technological progress are linked to HDTV through the ideograph of technological progress to suggest an 'ongoing evolution' of "a progress that should not be stopped" ( van Lente, 2000, pp. 47-48). Van Lente argues that HDTV became necessary because of the promises of technologists about what was feasible: The present system is deficient exactly to the degree that it does not live up to the projected characteristics of HDTV. The present is measured by the yardstick of the technological promise, and found wanting. As a result, the development is self-justifying ( van Lente, 2000, p. 56).
Underpinned by the ideograph of technological progress, the notion of a 'new generation' replacing an older generation becomes naturalised. Once the promise is made it leads to requirements -R&D activities and standardisation processes need to ensure that HDTV does indeed provide the promised improvements versus the existing technology (van Lente, 2000).

Technoscientific promises and the sociology of expectations
Anthony Giddens has argued that contemporary societies are more intensely future-oriented than has historically been the case (Giddens, 1997). Pierre-Benoit Joly adds that "[m]ost of today's social conflicts include a dimension related to the type of future we are creating through current choices and activities" (Joly, 2010, p. 2). For Joly, climate change is an archetypal example of an issue where "the future is so strong it determines the present" because widely believed representations of a catastrophic future shape present day behaviours aimed at avoiding such a future (Joly, 2010, p. 1).
At the turn of the century a number of scholars sought to investigate 1 Ideographs are understood as collectively held but vaguely defined normative goals that guide and justify action and belief in directions widely recognised to be acceptable (e.g. technological progress, democracy, freedom, equality, justice, independence, respect for the law, sustainable development, etc.) (McGee, 1980; van Lente, 2000).
this future-orientation through what became known as the 'sociology of expectations'. This involved 'looking at the future' -and in particular how the future is imagined, invoked and mobilised in the present in order to achieve particular effectsas an object of analytical focus, as opposed to 'looking into the future' (i.e. attempts to predict what will occur in the future) (Brown et al., 2000). In fact, from this perspective, attempts to look into the futurepredictions, statistical probabilities, forecasts, foresight and horizon scanning exercises, and so onand the media through which they are expressedgraphs depicting trend analyses, industry body reports, academic conferences, newspaper articles, science fiction, and so on -always have the potential to play a performative role in the present. In other words, expectations are not seen as merely attempts to anticipate a future that is unfolding independently of those attempts, but as (potentially) actively generating that future. As Brown and Michael put it, the sociology of expectations is the study of the "complex interactions between tools of prediction, discourses of the future and the shaping of the present" (Brown and Michael, 2003, p. 6). These scholars were particularly interested in the role that representations of the future play in processes of scientific and technological innovation. This is primarily because expectations play a particularly important role in the development of new technology, especially at the early stages of research and development. The reason for this is obvious policy and investment decisions are required about something which doesn't yet exist, and therefore must be made on the basis of future expectations rather than present realities.
Clearly there is a degree of similarity and interrelation (if not overlap) between a range of concepts that we would collectively term futureoriented discursive practices, or, put more simply, talking about the future. They include visions, promises, expectations, and imaginaries, alongside a range of broader discursive features that can but don't necessarily relate to the future, such as narratives, ideographs, storylines and frames.
In this paper, we will focus primarily on two of these closely related conceptsexpectations and promises. Van Lente defines expectations as "circulating representations of the future" (Van Lente, 2016, quoted in Sovacool, 2019, whilst Borup et al. define 'technological expectations' as "real-time representations of future technological situations and capabilities" (Borup et al., 2006, p. 286). Konrad et al. define expectations as "statements about future conditions or developments that imply assumptions about how likely these are supposed to be and that travel in a community or public space" (Konrad et al., 2017, p. 466). For Budde and Konrad, the inclusion of an implicit or explicit assessment of the likelihood of a state of affairs coming about in this definition is crucial to distinguish expectations from visionswhilst visions envisage a future they do not necessarily present a view about the likelihood of it coming about, whereas expectations do (Budde and Konrad, 2019). For this reason, in this paper we adopt Konrad et al.'s (2017) definition of an expectation.
Different types and levels of expectations have also been distinguished. For instance, expectations can be ascribed to particular individuals or groups, or be collective in nature. Collective expectations cannot be attributed to specific actors or groups but are instead "part of a generalised and taken-for-granted social repertoire" -it can be (and is) generally assumed that they are widely held or at least widely recognised (Konrad, 2006, p. 431).
Three different levels of expectations have generally been distinguished. The specific or project level concerns expectations about a specific manifestation of a technology or particular project, the general level concerns expectations about a technological field as a whole and the 'frame' level concerns societal hopes and fears concerning a technology (Kriechbaum et al., 2018;Ruef and Markard, 2010). The point here is not merely that expectations can be observed at different levels of technical specificity or societal generality, but that expectations at the different levels can interact in various ways. For instance, the work of scientists at the project level may be insulated from repercussions following disappointing results due to dynamics at the societal level (van Lente et al., 2013). As such, Budde and Konrad offer the concept of a 'network of expectations' to capture supportive and oppositional dynamic linkages between expectations across (vertical linkages) and within (horizontal linkages) different levels (Budde and Konrad, 2019).
Turning to promises, for Joly a promise starts with problematisation a statement to the effect of 'you have a problem and I have the solution' (Joly, 2010). Promises must then achieve legitimacy through widely acceptable 'good' reasons for favouring a particular option over the present situation and other possible options. In highlighting how a technology will solve problems and produce a future that is preferable to the present situation and other possible futures, promises seek to generate positive expectations about a particular favoured technological option. Credibility must also be secured, and given that technoscientific promises are inherently speculative, this typically relies upon the status of the proponents of a particular promise (e.g. experts, but also governments and firms with the capacity to invest in a particular technological option), as well as certain rhetorical strategies (e.g. presenting a technological development as natural) (Joly, 2010). In other words, legitimacy stems from reasons for pursuing a particular option that are widely held to be 'good', whereas credibility is achieved by all the factors that lead an audience to believe that a promise can and will in fact be delivered.
More broadly, there are various slightly different uses of the term promise in English. For instance, describing something as promising or demonstrating promise means that it possesses the quality of potential success or excellence. To suggest that technology A is a promising solution to problem B is, we would argue, tantamount to expressing a positive expectation about technology A. Where a promise clearly goes beyond a (positive) expectation is in the sense of making a promise. To make a promise is to make a more specific commitmentto enter into an informal contract. As Borup et al. put it, making a promise makes one accountable for doing something (Borup et al., 2006). Whilst expectations can create requirements for justification and action, a promise makes a more explicit commitment on the part of the promise-maker to actively ensure a particular future comes about. This distinction has something to do with the agency of the actor communicating an expectation or promisea promise-maker implies that they themselves have the capacity to deliver the promise (or at least play a role in doing so), whereas an actor setting out an expectation may do so on the basis of the perceived capacities of others to shape the future in the ways expected.
Another key difference is that whilst expectations can clearly be positive or negative; a promisein the specific, contractual sense of making a commitment to delivering a particular outcomeis rarely used in the negative, especially in the context scientific and technological innovation. It would be unusual for a scientist to commit to their work delivering a negative outcome. Therefore, by definition, a promisemaker views their favoured technological option as beneficial, whereas the expectations that this technology generates amongst wider audiences can be positive or negative. Whilst those that favour a particular technological option will make promises about what it will achieve or express positive expectations for it; where a number of competing candidate technological options exist, they will also tend to express negative expectations about rival options. Following Joly, we therefore define technoscientific promises as direct or explicit commitments to delivering a technological solution to a technical or societal problem, which imply the capacity to deliver on the part of the promisemaker.

The performative effects of promises and expectations
What then would motivate actors to articulate and amplify promises and expectations? The answer in part lies in the fact that promises and expectations can do effective work in the present that contributes to the achievement or avoidance of particular futures. Another way of saying this is that promises and expectations do not simply float in some abstract, a-material realm of language and representation (although they clearly have a discursive dimension); they take on material form in the shape of documents and images on a screen (Michael, 2000) and have concrete effects in the present in shaping technological development and choice. Van Lente summarises three key effects of expectationsexpectations (can) provide legitimation, direction and coordination for processes of innovation (van Lente, 2012). Legitimation refers to how expectations raise attention and legitimise investmentresources that are invested in emerging technologies and R&D projects can be justified on the basis of expected future performance. Direction refers to how expectations focus attention on particular technological options considered to be the most promising amongst the bewildering array of conceivable possibilities. Finally, coordination refers to how expectations structure research and innovation processes by highlighting tasks that need to be completed and roles that need to be performed.
More practically then, promises and expectations may capture the attention of scientific research funding bodies and R&D decision-makers at firms, and influence their funding decisions and R&D strategies. Promises and expectations also help to guide and coordinate the work of scientists, and may also result in the creation of a network of researchers striving to realise the promise of a particular technology (Bakker et al., 2011). Promises and expectations can legitimate the funding decisions of research councils, the work of researchers, and the strategic decisions of firms (van Lente, 2000); and grant researchers a mandate to pursue a particular line of inquiry (Bakker et al., 2011). They necessitate action on the part of those expressing themwhether through the requirement to justify particular expectations (e.g. through the communication of research findings) or through particular requirements that need to be met by promise-makers (e.g. results to be achieved within a certain timeframe) (Bakker et al., 2011;van Lente, 2000).
Promises and expectations can also influence government policy and regulatory and legislative reform where governments decide to pursue a particular technology because they have been convinced of its promise to help solve a particular problem (Cantoni et al., 2018). Furthermore, expectationsboth positive and negativecan mobilise publics to support or resist a particular technology, and civil society groups might adopt a particular expectation which will in turn influence their strategic activities (e.g. patients' groups lobbying governments to fund research into a promising new medical treatment) (Brown and Michael, 2003).
Widely held promises and expectations may become self-fulfilling prophecies, as in the case of Moore's law, where an expectation based on past trends became a yardstick against which actors measure required performance ( van Lente, 2012van Lente, , 2000. Expectations can also be self-justifying, whereby new generation technologies become necessary on the basis of expected performance improvements compared to the current generation, which researchers are then required to achieve (van Lente, 2000). Such dynamics can potentially lead to path dependencies, lock-in, and irreversibility (Borup et al., 2006;Joly, 2010). It is for this reason that Joly talks of "a sequence which flows from identification of [a] plurality of options to the making of a promise and then to requirements and necessity" (Joly, 2010, p. 10). Similarly, van Lente describes an option-promise-requirement-necessity sequence: [W]hat starts as an option can be labelled as a technical promise, and may subsequently function as a requirement to be achieved, and a necessity for technologists to work on, and for others to support (van Lente, 2000, p. 60). This is not an autonomous process, but the consequence of assessments of what is feasible, obsolete and necessary; and whilst the process is in principle reversible, the costs of doing so rise over time (and so the likelihood decreases) because of 'lock-in' dynamics (van Lente, 2000). However, it is perhaps just as likely that high expectations will be disappointed and promises will go unfulfilled as part of so-called hypedisappointment cycles (Borup et al., 2006;Joly, 2010). Actors are generally well aware of this latter possibility, often leading them to be sceptical of hype and big promises, and to expect disappointment (Brown and Michael, 2003).
In these and many other ways, promises and expectations can influence innovation trajectoriesthe particular technological options that become favoured and those that are abandoned or overlooked, and the way that the development of a technology unfolds. It is in this sense that many authors refer to promises and expectations as 'constitutive', 'generative' or 'performative' (Borup et al., 2006).
The above performative effects on practices, decisions and ultimately innovation trajectories are in part dependent on framing strategies that are employed to (attempt to) bolster the credibility of expectations in the eyes of target audiences. For instance, Bakker et al. investigate expectation credibility and argue that three elements are clearly visible in the assessments of technology selectors and the claims of technology enactors: a technologies current performance and its historical progress to that point, a path forward to higher levels of future performance, and an end-goal related to a perceived need or opportunity (Bakker et al., 2012). As we have already seen, successful promises are those that put forward widely accepted 'good' reasons for pursuing a particular technology (legitimacy) and those that audiences believe will in fact be delivered (credibility) (Joly, 2010).

Actors and activities involved in the production and circulation of promises and expectations
We finally turn to the actors that are involved in the production and communication of promises and expectations, and the roles that different groups are tasked with. 'Expectations work' (Bakker et al., 2011) or 'promissory work' (Pollock and Williams, 2010) is often done by scientists, industry actors, science journalists and technology consultants. Such actors can be thought of as 'brokers of futures' (Cantoni et al., 2018), and can often form into a network known as a 'technological community' (Bakker et al., 2011). Technological communities "are globally dispersed but have a shared goal to develop solutions and a shared interest in convincing technology selectors of the future potential of their work" (Bakker et al., 2011, p. 154). Bakker et al. (2011) make an analytical distinction between 'enactors' on the on hand, and 'selectors' on the other. The former research, develop and advocate for specific promising technological options, whereas the latter assess various competing options in order to identify which should be pursued or supported (Bakker et al., 2012(Bakker et al., , 2011. The distinction is analytical because the same actors can and do perform both roles, whether sequentially or even simultaneously (Bakker et al., 2012). For instance, firms that select a particular option will become enactors from that point, and enactors for a technological field may simultaneously act as selectors for particular technological components within that field (Bakker et al., 2012).
The promises and expectations of 'technological communities' tend to be aimed at audiences of potential 'selectors'; for instance, scientists, industry actors, policy-makers, research funders, and wider public audiences via the media and other engagement activities. 'Expectations work' involves producing and communicating expectations to such audiences, and challenging the expectations of rivals. More practically, this may involve conducting research, writing reports and articles, speaking at conferences and to the media, and so on. Such activities and sites can be thought of as 'arenas of expectations' in and through which trials of strength between rival technological communities and their competing expectations for different technological options are conducted (Bakker et al., 2011;Sovacool, 2019).
The following methodology section includes Table 2 which summarises how the conceptual work considered here provides a number key concepts which guided our analytical approach.

Method
We set out to analyse how the promise of 'green 5G' is produced and circulated. We asked: what are the particular contents of this promise (i. e. the more specific promises and expectations it is built upon), how do 'enactors' seek to boost its legitimacy and credibility, what activities are involved in its production and circulation and what are its present-day material effects, does the promise exclude or overlook alternative options, and is it challenged? As detailed in the following subsection, we pursue these aims and questions through a document analysis of a diverse corpus (n = 260) comprised of UK newspaper articles and newswire results, and reports and webpage content from industry, standardisation bodies and research consortia.

Data collection
Initial scoping searches of various possible 'arenas of expectations' led to the further exploration of some sites and document types, and the discounting of others. For instance, searches of Hansard, select committees and gov.uk revealed that formal sites of UK policy-making and legislation (e.g. parliament, government documents) were not arenas within which the promise of green 5G was being communicated and challenged. Expectations work relating to green 5G was however being conducted by research consortia, telecommunications companies, industry groups, and standardisation bodies; and being amplified and occasionally challenged through specialist trade press and to some extent the popular press as well. As such, we turned our attention to these 'arenas of expectations'.
We identified relevant documents in two different ways. First, we conducted a Nexis news search using the search string "5G" AND sustainab* OR green OR energy OR carbon OR "climate change" OR environment. We limited our results to the timeframe 01/01/10-15/07/21. We also limited our results to UK based sources in the English language for three particular document types: newswires, web-based publications and newspapers. The geographical scope of this part of the document corpus is therefore the UK (although the content of both newspaper and newswire results was not limited to events in the UK). Newswire documents provided more specialist material in the form of press releases and financial market memos (e.g. news of a new relevant report from a telecommunications company). On the other hand, newspaper results gave a sense of the stories that cut through to wider public audiences. We limited our newspaper results to national UK newspapers. Webbased publication results tended to cut across the previous two categories as they included both the online versions of newspaper articles as well as more specialist trade publications. As such, for simplicity, we included the former as 'newspaper' results and the latter as 'newswire' results.
We ran searches for newswire results and newspaper results separately (including web-based publications in both cases, and categorising online versions of newspaper stories as 'newspaper' results and more specialist results as 'newswire' results). Our newspaper search led to over 9000 results and our newswire search led to over 20,000 results. In both cases, we sorted the results by relevance and read the first 500 results of each to identify relevant documents. This approach resulted in the identification of 57 relevant newspaper results and 116 relevant newswire results. Both sets of results were subject to diminishing returns (i.e. there were a high number of relevant results on the first couple of pages of results and increasingly fewer relevant results in later pages), suggesting that sorting the results by relevance had successfully brought many of the most relevant results towards the beginning of the list of results.
In addition to the Nexis news database searches, we conducted manual internet searches for relevant research consortia, company, industry body and standardisation body documents and webpages. This involved compiling a list of relevant consortia, companies and bodies and using the same search terms as above. This approach led to the identification of 52 documents associated with companies and industry bodies, 27 documents associated with research consortia, and 8 documents associated with standardisation bodies. Our document corpus is summarised in Table 1.

Analytic approach
The document corpus described above was imported into Nvivo for analysis. The conceptual work covered in Section 2 provided a series of key concepts that formed the basis of our analytic framework, which is summarised in Table 2. This framework informed the development of a number codes which are aligned with the aims and questions expressed at the beginning of this section, and orientated and sensitised us to relevant and significant material within our corpus in the context of those aims and questions. Following a close reading of our documents, our codes were used to identify and classify relevant passages of the documents. The material assigned to each code was then subjected to another close reading in order to identify key themes and dynamics in response to our aims and questions, and to highlight the best illustrative quotes. We coded for positive expectations and promises and negative expectations, and notable themes within their content; strategies employed to boost the credibility and legitimacy of promises and expectations; and present-day performative effects of promises and expectations. Additionally, because it quickly became obvious that one of the performative effects of the promise of green 5G within our corpus was to focus attention on technologies and strategies aimed at rendering expected traffic growth sustainable (e.g. through energy efficiency improvements) rather than alternative approaches (e.g. strategies aimed at managing demand), we also coded statements about expected traffic growth and reasons for and proposed solutions to this problem.

Findings
In this section we present the findings from our analysis. We start by presenting positive expectations and promises that contribute to the promise of green 5G. We look at the different types of positive expectations put forward in our corpus and then discuss the strategies that green 5G 'enactors' employ to boost the legitimacy and credibility of their promises and expectations. We then cover negative expectations that raise doubts over or directly contest the promise of green 5G. Next, we discuss how different problematisations and framings position 5G in contrasting ways in relation to the issue of expected future traffic growth, and highlight a set of potential solutions that are almost entirely overlooked in our corpus. Finally, we consider a number of performative effects of the promise of green 5G.

Green 5G and greening through 5G
There are two distinct sets of positive expectations about 5G's potential to be a green technology regularly expressed in our corpus. The first focuses on the energy use or emissions of mobile networks themselves (with user devices receiving relatively little attention), whereas the second focuses on the potential for 5G to improve the efficiency of processes and reduce resource use and emissions in other areas of social and economic life. We'll refer to the former as expectations of green 5G and to the latter as expectations of greening through 5G.
Expectations of green 5G are voiced on the basis of two main approaches. Whilst some positive expectations focus on strategies aimed at improving the energy efficiency or reducing the energy use of 5G networks, another set of positive expectations relate to increasing the use of renewable energy sources to power networks. The latter are often not technological expectations relating to 5G itself, but often take the form of broader promises on the part of the industry to increase their use of renewable energy more generally. Positive expectations about these two strategies for green 5G are often expressed together.
The upgrade will make the network about five times more energyefficient than previous generations of mobile technology. In addition to being powered by 100 percent renewable electricity, this will deliver sustainability and power consumption benefits to Telia Lithuania (ENPNewswire, 2020).
A key source of expectations about 5G's ability to improve the energy efficiency of mobile networks comes from a Nokia and Telefonica press release, which states that "[a] new study by Nokia and Telefónica has found that 5G networks are up to 90 percent more energy efficient per traffic unit than legacy 4G networks" (FinancialWire, 2020). The press release goes on to describe 5G as "a natively greener technology with more data bits per kilowatt of energy than any previous wireless technology generation" (FinancialWire, 2020). Whilst the press release does acknowledge that "5G networks require further action to enhance energy efficiency and minimize CO2 emissions that will come with exponentially increased data traffic" (FinancialWire, 2020), the expected energy efficiency improvements are nonetheless used to justify describing 5G as inherently or naturally green. Both the study's headline finding and the designation of 5G as natively green are then repeated across a number of documents in our corpus.
Expectations of greening through 5G sit within the context of broader expectations about the decarbonisation enabled through ICTs and digitisation in general. The Global e-Sustainability Initiative (GeSI), a strategic partnership of ICT companies, is a key proponent of such broader expectations about the role digital technologies can play in decarbonisation (GeSI, 2015) and achieving the UN's sustainable development goals (GeSI, 2019). These broader expectations about greening through digital and smart technologies are then picked up and applied to 5G specifically in many of the documents in our corpus. For example: 5G will drive industries and communities towards digitalization, automation and advanced analytics. For instance, a 5G-enabled factory can have more environmentally sustainable and efficient processes through predictive maintenance, on-demand manufacturing and use of smart logistics (ENPNewswire, 2019).

Energy efficiency or energy use
In terms of specific outcomes expected of 'green 5G', some actors merely express positive expectations about increasing the energy efficiency of networks (e.g. the Nokia example above), whereas others go further to express expectations of network energy use remaining flat or falling compared to today's levels of energy use in spite of expected dramatic increases in data traffic.
Another example of expected energy efficiency improvements being used to bolster the promise of green 5G comes from Orange: Because it will be 10 times more energy efficient than 4G by 2025, 5G and digital will make a positive contribution to the environmental transition of businesses and society (Orange, 2021).
However, expectations of 5G energy efficiency improvementsreducing how many watts it takes to transmit a bitare arguably not particularly compelling reasons to consider 5G as green. This is because dramatic increases in data traffic could result in an overall increase in network energy consumption in spite of the efficiency gains. A number of the documents in our corpus acknowledge this. For instance: This growing energy challenge is, in large measure, a result of the exponential growth in traffic that new 5G services are likely to deliver. Although the 5G-new-radio standard is more energy efficient per gigabyte than are the 4G standards, the proposed 5G use cases and new spectrum bands will require many more mobile sites, outstripping potential energy efficiencies (McKinsey and Company, 2020, p. 2).
At times in our corpus there is significant ambiguity about the specific levels of performance or particular achievements that are expected of green 5G. This can arise due to a lack of clarity over baselines when comparing expected reductions in energy use or emissions, a lack of detail about expected traffic levels in scenarios, and ambiguity about the overall impact on total network energy consumption implied by various expectations. In these ways, the promise of green 5G is to some extent and at sometimes circulated and maintained through vague expectations that are open to misinterpretation. Take the following example: In addition, 5G offers significantly enhanced energy efficiency. The new generation of mobile phones consumes up to 90 percent less energy per byte than 4G. CO2 emissions are also reduced by a Technological solutions to problems (technical or broader societal) (Joly, 2010) An explicit and direct commitment to delivering a particular outcome, implying the capacity to deliver on the part of the promise-maker Elements of successful promises and expectations Positive assessments of a technology's current performance and its historical progress to that point, a path forward to higher levels of future performance, and/or an end-goal related to a perceived need that the technology is expected to meet or opportunity that the technology is perceived to present (Bakker et al., 2012) Highlighting widely acceptable 'good' reasons for favouring a particular option over the present situation and other possible options (to boost legitimacy) (Joly, 2010) Highlighting factors that lead an audience to believe that a promise can and will in fact be delivered (to boost credibility) (Joly, 2010) Performative effects of promises and expectations The creation and circulation of texts and images ( Michael, 2000) Directing, legitimising and coordinating research agendas and activities and innovation processes ( van Lente, 2012) Requirements for actionto justify and meet expectations and deliver promises (Bakker et al., 2011;Joly, 2010; van Lente, 2000) Closure towards particular innovation trajectories, curtailing or overlooking alternative pathways (Borup et al., 2006;Joly, 2010;van Lente, 2000) significant amount. Data transmission of one hour of video streaming over 5G causes results [sic] in only 5 percent of the CO2 emissions of such streaming over the legacy mobile phone standard 3G (Market-Line NewsWire, 2020).
This statement focuses on the emissions associated with a particular service (i.e. 1 h of video streaming) and so neglects the possibility that increases in the use of that service will counteract efficiency improvements. As such, the statement that 'CO2 emissions are also reduced by a significant amount' may encourage the potentially erroneous expectation that 5G will reduce CO2 emissions at the network level.
At other times, however, the expected results of green 5G are more ambitious and well-defined. For instance, in their 'Breaking the energy curve' report, Ericsson set out the expectation that if 5G deployments follow their 'breaking the energy curve' approach, then data traffic can be quadrupled without increasing network energy consumption (Ericsson, 2020). Similarly, Nokia argue that by using a range of approaches (e.g. sleep modes, improved hardware efficiency, small cells) overall network power consumption "has the potential to remain flat between 2015 and 2025" in spite of heavy traffic growth (Nokia, 2016, p. 11).

Legitimacy and credibility
The legitimacy of the promise of green 5G is established by positioning 5G as the solution to the problem of expected growth in mobile data traffic. The anticipated growth in data traffic is regularly referred to as exponential within our corpus. Green 5G is widely seen as essential to serving these increases in data traffic in an environmentally and economically sustainable fashion.
Why do we need 5G? Mobile data traffic is rising rapidly, mostly due to video streaming. With multiple devices, each user has a growing number of connections. Internet of Things will require networks that must handle billions more devices. With a growing number of mobiles and increased data traffic both mobiles and networks need to increase energy efficiency. Network operators are under pressure to reduce operational expenditure, as users get used to flat rate tariffs and don't wish to pay more (ETSI, 2021).
Within this framing of the problem though, there are subtle differences. As already discussed, some accounts see 5G as 'natively green' and therefore view 5G as the solution to the problem of the future energy use of mobile networks in the context of increasing data traffic. For example, a report by STL Partners voices the expectation that 5G will be more energy efficient than 4G. The report expects that a 5G mmWave site has the potential to take just 1-2 % of the energy it takes a 4G site to transmit a unit of data (STL Partners, 2019). As such, the report argues that the scope for telcos to reduce energy use and emissions depends on how quickly they roll out 5G -the quicker they roll out 5G, the more energy and emissions they'll save between 2018 and 2030 (STL Partners, 2019). Note that the requirement for action that stems from this problematisation is to roll 5G out as quickly and as widely as possible.
Others, however, don't see 5G as inherently green, but rather as a mobile network generation that could be made to be more efficient or greener, and could be deployed in a more or less sustainable way. In this framing, greener versions of 5G are the solution to anticipated increases in data traffic. The requirements for action that stem from this problematisation are of course to conduct research and innovate to improve the energy efficiency of 5G and to deploy existing and new energy efficiency solutions such that increases in data traffic are rendered sustainable. Generally, within our corpus, there is a good deal of optimism that anticipated traffic increases can be rendered sustainable through strategies and technologies aimed at improving the energy efficiency of future mobile networks (alongside other measures such as purchasing renewable electricity).
As Mats Pellbäck Scharp, Head of Sustainability at Ericsson, stated at the conference, it is important that we break the rising energy curve. This is doable with 5G as it is 90 % more energy efficient compared with 4G, but it also requires actions such as replacing inefficient network equipment and deploying energy saving functionalities (GSMA, 2020).
Those voicing promises and expectations about green 5G rely on a range of sources of credibility. The most widely used of these sources of credibility is to highlight both specific technical advances compared with previous mobile network generations and broader approaches and strategies aimed at improving energy efficiency and increasing the use of renewable energy. Because 5G started to be rolled out during our timeframe, such advances are often presented as having already been achieved and demonstrated, and the results of trails and early deployments are a further important source of credibility. Alongside such early real-world results; models, simulations and forecasts of the potential future efficiency improvements or emissions and energy use reductions made possible by 5G remain an important source of credibility. Such specific predicted impacts are also regularly joined by more generic expressions of future technological potential (e.g. the increasing use of AI for managing network energy use). Finally, as well as establishing credibility by looking to the future, in our corpus 5G 'enactors' also point to the telecommunications industry's historical track record of improving the energy efficiency of mobile networks with each generation and successfully implementing more specific green initiatives.

Negative expectations
Whilst positive expectations about green 5G dominated our corpus, we also identified a number of contrary expectations. We identified 625 instances of positive expectations concerning the promise of green 5G being expressed within our corpus, compared to 111 negative expectations.
Negative expectations are sometimes expressed with the aim of mobilising research and innovation efforts to ensure that 5G is ultimately green. For instance: The potential increase in data traffic (up to 1,000 times more) and the infrastructure to cope with it in the 5G era could make 5G to [sic], arguably, consume up to 2-3 times as much energy. This potential increase in energy, coming from a high number of base stations, retail stores and office space, maintaining legacy plus 5G networks and the increasing cost of energy supplycall for action (GSMA, 2019). This negative future is therefore mobilised in order to encourage and justify present-day actions that will prevent its occurrence. Such expectations therefore aim to be 'self-preventing' -a particular representation of the future is expressed with the aim of preventing its occurrence. At other times, however, negative expectations are more straightforwardly predictions of what will occur. Exact estimates differ by source, but MTN says the industry consensus is that 5G will double to triple energy consumption for mobile operators, once networks scale (Fierce Wireless, 2020).
Negative expectations are based on the view that 5G will generate more data traffic and more data-intensive services which will in turn require network densification (i.e. more base stations). Whilst negative expectations acknowledge that 5G is more energy efficient than previous generations, they are based on the view that efficiency improvements will be insufficient to offset the energy demands of traffic increases and densification.
Although 5G networks are up to 90 percent more energy efficient than legacy 4G networks, overall energy usage will increase due to rising traffic and the rollout of ever more dense networks (M2 PressWIRE, 2020).

Problematisation, framing responsibility for traffic growth, and overlooked alternatives
As we have seen, expectations of dramatic increases in mobile data traffic are a key concern for proponents and opponents of the promise of green 5G alike. A crucial feature of discourse within our corpus is therefore how actors both position 5G in relation to the problem of expected traffic growth and assign responsibility for this growth. Another crucial feature of discourse within our corpus is how processes of problematisation focus attention on certain solutions at the expense of others. Three key questions that arise here are: a) what causes the expected growth in data traffic, b) what is the solution to this problem, and c) do problematisation processes within our corpus direct attention to particular solutions and away from others? Some sources within our corpus implicitly frame expected data demand increases as being independent from 5G. As such, 5G becomes a necessary response in order to render such increases in data traffic sustainable. For instance, the STL Partners report suggests that: [T]he inexorable growth in data volumes over mobile networks means that relying on "evolutionary" improvements to 4G will not be sufficient to curtail emissions from networks (STL Partners, 2019, p. 3).
Followed by: Network operators should see 5G network deployment not only as an investment in their services, but also as part of a strategy to contain growing energy demands of escalating data volumes (STL Partners, 2019, p. 5).
This framing of the problem implicitly suggests that 5G is not responsible for expected data traffic increases but is the solution to rendering such increases sustainable. The requirement for action that stems from this problematisation is to deploy 5G as quickly and widely as possible.
On the other hand, other actors quite explicitly suggest that 5G will itself be at least partly responsible for expected traffic growth.
The likelihood is that, just as 4G got us all using more data, watching TV shows on our phones or tablets, 5G looks set to drive another dramatic expansion in data demand (The Telegraph, 2019).
In this framing, 5G in part causes the problem to which it is presented as being the solution (i.e. dramatic traffic growth). The dominant view in our corpus however is that 5G efficiency improvements can indeed counteract traffic growth and its energy use implications and so render mobile networks sustainable. The requirement for action that stems from this framing is to ensure that 5G energy efficiency improvements and other measures (e.g. renewable electricity purchasing) are sufficient to render the data traffic growth that is at least partly caused by 5G sustainable.
The final point to raise in this subsection is that a set of possible solutions to the problem of the energy use implications of expected traffic growth is conspicuous by its almost complete absence in our corpus. Such solutions might stem from a problematisation that doubts whether 5G energy efficiency improvements will be sufficient to offset the traffic growth that 5G itself partially causes. Here we are referring to the prospect of demand-side management. That is, approaches and strategies aimed at reducing or modifying consumer and business demand for mobile data, rather than efforts aimed towards catering for demand in more efficient and sustainable ways. We did find a small number of references (n = 10) in our corpus that floated the idea of limiting or managing demand for data in some way. A first, rather uncontroversial step to managing demand is to increase consumer awareness of the environmental impact of data use with a view to encouraging more conscious, less energy-intensive practices.
[W]ith regard to the increase in traffic expected over the same period, necessary educational and support activities related to customer usage will also have to be carried out to move toward optimal energy conservation (Orange, 2020).
Of course, Orange present such activities as being complementary to 5G energy efficiency technologies, but the implication is nonetheless that customers might have to be informed about the impacts of their usage in order to encourage greater energy conservation. It is worth noting that this framing implies that customers are at least partially responsible for addressing the environmental consequences of their data use.
Encouraging less energy-intensive practices doesn't necessarily involve reducing demand, but could rather involve nudging demand towards more efficient access networks (e.g. WiFi rather than 4G).
Dr. Bitterlin added: 'If you're doing a heart operation somewhere, with remote video, then absolutely yes [you need HD]. 'Watching a video on a phone, with most people's average eyesight? Absolutely not.' Dr. Hazas pointed out that streaming a video or doing anything else on the internet also uses three to four times more energy if you're on 3G or 4G compared to if you're on WiFi. 'If you have the choice, watching that video at home as opposed to during your daily commute is a better way,' he advised. 'In terms of energy consumption, things like autoplay and infinite scrolling, because they keep us engaged, that keeps the data flowing which keeps the energy flowing. 'I've got a three-year-old and when he's a teenager, I would hope that we'd also think about, "Oh is this a video that I really wanted to stream, am I really going to be here in the next five minutes to actually watch it, or is this the kind of thing I should switch off?"' (MailOnline, 2020).
Other interesting elements of the above passage include the suggestion that certain energy-intensive services (i.e. HD streaming) serve no appreciable purpose in many contexts, as well as the suggestion that some digital services are designed to stimulate traffic and encourage thoughtless consumption (in contrast to the point above about consumer responsibility).
Statements that suggest that demand should be limited or managed in some way are in a tiny minority in our corpus. The overwhelmingly dominant perspective is that mobile data enables useful services, growing demand for mobile data should be satisfied, and technological innovation can ensure that 5G will do this in a sustainable manner. Promises and expectations within our corpus overwhelmingly shift attention towards technological options and strategies designed to serve increasing demand for data in sustainable ways (e.g. by improving the energy efficiency of networks and purchasing renewable electricity) and away from approaches that seek to manage demand in some way. In this way, the promise of green 5G can be seen to have the performative effect of shaping innovation trajectories towards particular technological options and away from others.

Performative effects
As discussed in Section 2, promises and expectations take on material form in the form of text and images and (can) have material effects through influencing research agendas, government policy, public opinion and innovation trajectories. Our corpus itself is a reservoir of text and images that embody and articulate (and occasionally challenge) the promise of green 5G. Our corpus also highlights a number of activities, initiatives and decisions that are both driven by and aim to pursue the promise of green 5G. These effects are also driven by the threat of the undesirable future of mobile networks that are not green or sufficiently green (in the sense of self-preventing expectations discussed previously).
Beyond the performative effect of directing innovation trajectories already discussed, other performative effects evident within our corpus include the funding of research consortia aimed at developing energy efficiency strategies and technologies, academic and industry conferences and workshops focusing on the promise of green 5G, the environmental goals and commitments of telecommunications companies, tests and trials aimed at demonstrating the promise of green 5G in realworld environments, and the co-ordinating efforts of standardisation and industry bodies on environmental matters.

Discussion
In the first part of our discussion, we summarise the empirical insights that emerge from the analysis presented in the previous section. We also briefly discuss some of the ways that our findings are consistent with or differ from the findings of the studies discussed in Section 2.1. In the following subsection, we propose two broader conceptual insights that emerge from this study.

Empirical insights
A key contribution of this paper is the empirical contribution of shedding light on how promises and expectations operate in the case of the sustainability implications of 5G. As discussed in Section 2 above, these concepts have notto the best of our knowledgebeen applied to this empirical case before. Our analysis of a diverse corpus of documents reveals a number of different types of promises and positive expectations that contribute to the production and circulation of the promise of green 5G. This includes promises and expectations concerning the energy use or emissions of networks themselves (which we termed expectations of 'green 5G'). 2 Another set of positive expectations concern the potential for 5G to enable energy use and emissions reductions in other areas of social and economic life (which we termed expectations of 'greening through 5G'). Within the set of expectations of green 5G, we identified a further distinction between expectations of energy efficiency improvements compared to existing networks and those that envisage overall network energy use or emissions remaining flat or falling as a consequence of the deployment of 5G. At times in our corpus, there was significant ambiguity about the impact of 5G on overall network energy use or emissions, and the basis on which 5G should be considered green.
The legitimacy of the promise of green 5G was boosted by presenting 5G as the solution to the problem of rendering expected traffic growth sustainable. Some expectations saw 5G as inherently greener that 4G on the basis of energy efficiency improvements, leading to the requirement to roll 5G out as quickly and widely as possible. Other expectations saw 5G as a technology that could be more or less green or deployed in a more or less sustainable fashion, leading to the requirements to conduct research to further improve the efficiency of 5G and to deploy existing and new efficiency technologies and strategies to ensure that 5G was as green as possible. In general, within our corpus there was a good deal of optimism that 5G would render anticipated traffic growth sustainable. The credibility of the promise of green 5G was boosted by pointing to the track record and historical progress of the telecommunications industry, emphasising present day technological capabilities and available strategies and their demonstration through trails and early deployments, and highlighting forecasts of the future improvements that are possible and the more generic scope for further technological improvements to come.
Whilst positive expectations dominated our corpus, they were nonetheless contested by a significant minority of negative expectations. Some negative expectations were expressed with the aim of being selfpreventingthe possibility of an undesirable future was expressed to mobilise the present-day action required to avoid that future. Other negative expectations were more straightforwardly predictions of what would occur, and were based on the view that energy efficiency improvements would be insufficient to offset the energy use impacts of anticipated traffic increases.
We also highlighted the importance of the various ways that 5G was positioned in relation to mobile data traffic growth as part of processes of problematisation and how responsibility for this traffic growth was framed, and noted that a set of possible solutions to the problem of this growth was almost entirely overlooked in our corpus. Some accounts in our corpus imply that traffic growth is independent of 5G and suggest that 5G is 'natively green'. The solution to the problem of the energy use implications of traffic growth is therefore to roll out 5G as quickly and widely as possible. Other accounts acknowledge that 5G will at least in part cause problematic traffic growth, but nonetheless express optimism that efficiency improvements can render such growth sustainable. The implication for action of this problematisation is to take steps (i.e. researching further improvements, ensuring available solutions are deployed) to ensure that 5G is efficient enough to do so. An alternative problematisation that is rarely fully articulated in our corpus starts with negative expectations about the ability of efficiency improvements to render traffic growth sustainable. This leads to the problematisation of the dominant approach of stimulating and satisfying demand in sustainable ways and suggests that demand management solutions may be necessary. Statements that raise such solutions are in a tiny minority in our corpus, suggesting that the promise of green 5G is directing focus towards certain possible solutions and away from others.
Finally, we summarised some of the key performative effects of the promise of green 5G that were evident within our corpus. Such effects were driven by and sought to achieve the promise of green 5G and wider commitments to ensure the sustainability of future mobile networks. They included funding, legitimising and guiding the work of research consortia; telcos committing to environmental targets; the establishment of test and trial programmes; and the coordinating efforts of industry and standardisation bodies.
Given the lack of work focusing on the emerging, dominant, societallevel expectation of green 5G (which we term the promise of green 5G), the empirical insights summarised in this section represent a valuable empirical contribution. Beyond this, our findings demonstrate some synergies with the work of Mansell. We identify similar dynamics to those emphasized by Mansell in relation to issues such as surveillance, privacy and data protection (Mansell, 2021;Mansell and Plantin, 2020). In debates about the sustainability implications of 5G, in our corpus at least, we find that dominant expectations are directing focus towards particular solutions to the problem of the energy use implications of traffic growth and away from alternatives.
In relation to van Lente's arguments about HDTV (van Lente, 2000), we certainly see the same naturalisation of ongoing evolution expressed in our case through the generational naming convention for mobile networks. However, as explored in the following subsection, whilst we do detect the same self-justifying dynamic at work for other more general promises of 5G (e.g. the promise of enhanced mobile broadband), we argue that the promise of green 5G becomes necessary through a different process.

The necessity of green 5G and the role of environmental discourses
Our analysis of the case of the promise of green 5G also offers two of broader conceptual contributions concerning how the promise of green 5G comes to be seen as necessary and why actors adopt positive expectations regarding the scope for technological innovation to render growth sustainable.
In relation to van Lente's work (van Lente, 2000) on the selfjustifying dynamic brought about by the ideograph of technological 2 It is worth reiterating here that promises and positive expectations concerning the energy use or emissions associated with user devices were much less common within our corpus. This may simply be a consequence of framing our research around 5G, or it may alternatively suggest that solutions aimed at improving the sustainability of user devices are relatively overlooked. progress (whereby new generations of a technology can become necessary on the basis of promises about what is feasible above and beyond the capabilities of present generations), we see a slightly different dynamic in this case having to do with the way that competing ideographs pull in different directions. It seems evident to us that many promises about 5G in general do indeed operate in the way van Lente describesfor instance, faster data download speeds than 4G are established as feasible and promised which then necessitates their achievement due to widespread commitment to the ideograph of technological progress. We would argue, however, that the promise of green 5G is not established as necessary merely because it is feasible, but because it is required by a different ideographthat of sustainable development. So whilst general performance improvements do indeed become required through the logic of 'ongoing evolution' identified by van Lente, these performance improvements have energy use implications as they encourage dataintensive services and growth in data traffic. In turn, this necessitates action to render this new generation of technology sustainable. As such, the consequences of commitment to one ideographtechnological progresscreates a requirement for action in order to avoid violating a commitment to another ideographsustainable development. The commitment to technological progress also creates a general sense of optimism that the action required to render 5G sustainable (primarily through technological innovation) can indeed succeed. In this way, the promise of green 5G can be seen as a necessary response to the dual commitment of the telecommunications technological community to both the ideographs of technological progress and sustainable development.
This seems to us to have implications for the 'option-promiserequirement-necessity' sequence (Joly, 2010;van Lente, 2000). The promise of green 5G does not become necessary because promises and expectations establish it as feasible, but because the dual commitment to both technological progress and sustainable development demands that traffic growth is stimulated and satisfied and that this is done in a sustainable way. So whilst some general promises of 5G do begin from assessments of future capabilities that are feasible that go beyond the capabilities of 4G (e.g. faster download speeds), given that these future capabilities must be achieved in a sustainable way, it is arguably these twin necessities that kick-start the production, circulation and pursuit of the promise of green 5G. There is also an implication here for processes of problematisation. The solutions that are pursued are those that respond to the problem of 'how to render the necessary (due to the commitment to technological progress) next generation sustainable', as opposed to the rather broader problem framing of 'how to render mobile networks sustainable'. The latter problem framing would lead to the consideration of a broader range of possible options, including those that violate the necessity of delivering advanced capabilities that are technologically feasible but potentially challenging from a sustainability perspective.
The final point we discuss in this section concerns the question of why certain actors are drawn to and take up particular promises and expectations. A clear explanation provided by the sociology of expectations for why certain actors will adopt particular promises and expectations is that they can result in desirable performative effects in the present. A strong implication here is that such effects can often be in the interests of certain actors. For instance, it is in the interests of scientists working in a particular field to be granted further research funding on the basis of positive expectations about the field. We can see similar dynamics in our case, however we would argue that our case suggests an additional possible explanation for why certain actors are drawn to the promise of green 5G.
In our corpus we find a dominant optimism that technological innovation in the form of network energy efficiency improvements can be delivered at sufficient levels and pace to offset the energy use implications of expected growth in data traffic and so render such growth sustainable. These positive expectations are, we argue, strongly associated with particular environmental discourses. For Dryzek, discourses are most simply shared ways of apprehending the world (Dryzek, 2013). Discourses enable those that subscribe to them to "interpret bits of information and put them together into coherent stories or accounts" (Dryzek, 2013, p. 9). Each discourse rests on particular assumptions, judgements and contentions; and they help to define common sense and legitimate knowledge (Dryzek, 2013). They are embedded in language but can also embody power (e.g. by conditioning perceptions and values) and be constrained by material political realities (e.g. the fact that Governments feel they need to ensure economic growth). Dryzek identifies four overarching, competing environmental discourses (and a number of variants within these basic types) that have dominated recent environmental politics (Dryzek, 2013). Two of these environmental discourses are dominant within our corpus.
Ecological modernisation is a variant of the overarching discourse of sustainability. Ecological modernisation advocates for interventions aimed at reforming but not radically reimagining "the capitalist political economy along more environmentally sound lines" (Dryzek, 2013, p. 170). A key feature of ecological modernisation is that industry cooperates with such reform because it is in their interests to do so. This is because, amongst various reasons, pollution is inefficient; prevention is less costly than environmental restoration; and there are commercial opportunities in green technologies, goods and services. Ultimately, ecological modernisation seeks to decouple economic growth from environmental damage, as captured in the notion of 'green growth'. With economic and environmental values no longer in tension, actors that might previously have seen themselves as in opposition to environmentalism (due to the negative impact of environmental action on economic growth) are now free to become environmentalists. In dissolving what had previously been considered a trade-off between economic growth and environmental protection, ecological modernisation suggests that societies can have both in a 'win-win' scenario. It is for this reason that Dryzek describes ecological modernisation as a 'discourse of reassurance' (Dryzek, 2013, p. 175).
Because technology tends to be key to efforts to decouple economic growth from environmental damage 3 and because its 'win-win' denial of a necessary trade-off between growth and environmentalism implies a certain hedonism (and a corollary rejection of the austerity of limits), ecological modernisation shares certain similarities (amongst many differences) with another environmental discoursethat of 'Promethean environmentalism'. Prometheans exhibit "unlimited confidence in the ability of humans and their technologies to overcome" environmental problems (Dryzek, 2013, p. 52). Whilst Promethean environmentalists recognise the severity of environmental issues (unlike other types of Prometheans), they believe they can be overcome by technologies developed and deployed by both governments and markets. Promethean environmentalists envisage a future of clean, green abundance; secured through human ingenuity and technological innovation. They therefore have an almost boundless commitment to what we have referred to as the ideograph of technological progress and believe that environmental problems can be addressed without the need for limits on or changes to resource-intensive behaviours, because such behaviours can be rendered sustainable through technological innovation.
Ecological modernist and Promethean environmentalist tendencies are very clearly evident within our corpus. Such tendencies are evident in acceptance that improving efficiency and reducing waste are desirable aims in the interests of companies themselves, high levels of confidence in the ability of technological innovation to render traffic growth sustainable, and the almost complete absence of discussion about behaviour change and demand management.
More generally, limitlessness is something of a motif of 5G. It features in relation to 5G in customer-facing advertising campaigns in the context of the launch of unlimited data plans (BeUnlimited -Vodafone, 2019), corporate visions (Limitless Connectivity -Ericsson, 2021b), industry body blog posts (limitless innovation -5G Americas, 2021) and academic research institute press releases (limitless possibilities -5GIC, 2016). Koomey's lawreferred to in the title of this paperis not a natural law (as its name potentially implies), but the result of trend analyses of the energy efficiency of computing. In 2011, Koomey et al. reported that between 1946 and2009 the energy efficiency of computation at peak output doubled every 1.57 years (Koomey et al., 2011). Koomey and Naffziger later observed that between 2000 and 2009 the rate of doubling slowed to about every 2.6 years (Koomey and Naffziger, 2016). Whilst Koomey certainly did not predict limitless improvements in the energy efficiency of computing, Koomey's law and the closely associated Moore's law (Greene, 2011) have nonetheless both become bywords and short-hand justifications for optimism about the scope for innovation to improve the energy efficiency of ICTs. For instance, note how Moore's law provides inspiration for the notion of 'exponential technologies' -those that see rapid increases in performance as costs fall and the role they can play in meeting climate change targets (Falk and Gaffney, 2019). Positive expectations about the scope for efficiency improvements to go on rendering growth sustainable underpinned by the emblem of Koomey's law, the resulting lack of attention paid to demand management, and the wider motif of limitlessness suggest that the promise of green 5G resonates strongly with ecological modernist and Promethean environmentalist perspectives. We therefore suggest that commitments to underlying environmental discourses can help to explain why certain actors are convinced by, drawn to and take up particular promises and expectations about the sustainability implications of an emerging technology.

Conclusion
The emerging promise of green 5G is established and maintained through a range of more specific promises and expectations concerning 5G's impact on the energy efficiency, energy consumption and emissions of mobile networks; and the scope for 5G to enable efficiencies and savings in other areas of economic and social life. The legitimacy of the promise of green 5G rests on the view that 5G will help to render expected traffic growth sustainable; and its credibility is boosted through appeals to the track record and historical achievements of the telecoms industry, current technological capabilities, and further expected improvements and future potential. Some negative expectations seek to achieve the promise of green 5G by acting as self-preventing expectations, whereas others contest the promise green 5G by expressing doubts about the scope for efficiency improvements to counteract the energy use implications of traffic growth. The promise of green 5G has performative effects in legitimising and coordinating research and innovation efforts. Perhaps most importantly, the promise of green 5G directs attention towards approaches that aim to render expected traffic growth sustainable through energy efficiency improvements and away from other potential solutions based on demand management.
The promise of green 5G emerges as a necessary response to the telecommunications industry's dual commitment to technological progress and sustainable development. The former demands that feasible advances beyond the technical capabilities of 4G are necessary, whereas the latter demands that such advances must be achieved in a sustainable way. Confidence that these two demands can be simultaneously metthat 5G can render traffic growth sustainableexpresses and stems from the adoption of the environmental discourses of ecological modernisation and Promethean environmentalism. The role that Koomey's law plays as a byword and short-hand justification for such confidence is emblematic of these underlying environmental discourses. These modes of environmental thinking encourage the view that growth (in this case mobile data traffic growth) can be decoupled from environmental damage through technological innovation, and so largely reject the need for limiting or changing resource-intensive behaviours.

Funding
This work was supported by UK Research and Innovation, through a grant to the Centre for Research on Energy Demand Solutions (grant number EP/R035288/1), and the Leverhulme Trust, through a Leverhulme Early Career Fellowship (ECF-2020-405).

Data availability
The data underpinning this study are openly available from the UK Data Service's ReShare repository at https://doi. org/10.5255/UKDA-SN-856135.

Dr. Laurence Williams is a research fellow in environmental politics at the Science Policy
Research Unit (SPRU) at the University of Sussex, UK. His work focuses on sociotechnical controversies; discursive contests surrounding new technologies; and public perceptions of and engagement with science, technology and innovation. He has explored these themes through engaging with two cases in particularhydraulic fracturing and 5G mobile networks.
Dr. Noam Bergman is a lecturer in energy policy at the Science Policy Research Unit (SPRU) at the University of Sussex, UK. His research history includes both natural and social science work. His background is in natural sciences, with a BSc in Physics and an MSc in Environmental Sciences, both from the Hebrew University, Jerusalem, and a PhD in environmental sciences from the University of East Anglia (UEA). Since 2005, his research has focused on transitions to sustainability. This has included researching transitions to sustainable development as part of the EU project MATISSE; and studying microgeneration from a socio-technical perspective and local, community based action on climate change and climate activism whilst at the University of Oxford's Environmental Change Institute (ECI). Since early 2015 he has worked at SPRU, first as a researcher at the Centre on Innovation and Energy Demand (CIED), and now as a lecturer in Energy Policy.