Energy efficiency potentials in the EU industry: impacts of deep decarbonization technologies

Increasing the energy efficiency in high energy demand sectors such as industry with a high reliance on coal, oil and natural gas is considered a pivotal step towards reducing greenhouse gas emissions and meeting the Paris Agreement targets. The European Commission published final energy demand projections for industry capturing current policies and market trends up to 2050. This Reference scenario for industry in 2050, however, does not give insights into the extent to which energy efficiency potentials are already implemented, in which sectors further efficiency can be achieved, to what extent or with which technologies. In this paper, the EU Reference scenario is broken down and compared to a Frozen Efficiency scenario with similar GDP developments but without energy efficiency. Through bottom-up analyses, it is found that with energy efficiency technologies alone, this Reference scenario for industry energy demands (10.6 EJ in 2050) cannot be achieved. That means that the EU Reference assumes higher energy efficiency than possible and too high an effect of current policies. In the Frozen Efficiency scenario, the energy demand reaches 14.2 EJ in 2050 due to the GDP development; 22% higher than 2015. Energy efficiency improvements and increased recycling can decrease industrial energy demand by 23% (11.3 EJ in 2050). In order to further reduce the energy demand, our analyses shows that the wide implementation of innovative in combination with electrification or hydrogen technologies can further decrease the 2050 energy demand to 9.7 EJ or 10.3 EJ, respectively.

the extent to which energy efficiency potentials are already implemented, in which sectors further efficiency can be achieved, to what extent or with which technologies. In this paper, the EU Reference scenario is broken down and compared to a Frozen Efficiency scenario with similar GDP developments but without energy efficiency. Through bottom-up analyses, it is found that with energy efficiency technologies alone, this Reference scenario for industry energy demands (10.6 EJ in 2050) cannot be achieved. That means that the EU Reference assumes higher energy efficiency than possible and too high an effect of current policies. In the Frozen Efficiency scenario, the energy demand reaches 14.2 EJ in 2050 due to the GDP development; 22% higher than 2015. Energy efficiency improvements and increased recycling can decrease industrial energy demand by 23% (11.3 EJ in 2050). In order to further reduce the energy demand, our analyses shows that the wide implementation of innovative in combination with electrification or hydrogen technologies can further decrease the 2050 energy demand to 9.7 EJ or 10.3 EJ, respectively.

Introduction
The industrial sector is the single largest CO 2 emitter in the world. In 2018, the CO 2 emissions released from fuel combusted to drive industrial activities reached 14.4 Gtonnes (Gt), with 54% coming from direct fuel combustion and 46% from indirect fuel combustion for the generation of electricity (International Energy Agency [IEA], 2020). In the same year, the industrial final energy demand reached 164 EJ, 1 75% of which came from burning fuels and the rest from electricity use (IEA, 2020).
Future material demand is a determining factor for the development of industrial energy demand and related greenhouse gas emissions (GHGs). According to IEA (2019), global steel, cement and aluminium demand is expected to grow in 2060 by 30%, 10% and 75% (compared to 2017), respectively. The World Energy Outlook (WEO) (IEA, 2021) estimates that if current trends continue, total final energy consumption will increase by about 33% (from 156 EJ in 2020 to 207 EJ in 2050). To meet global climate targets and limit the temperature to 1.5 °C, IEA recommends that CO 2 emissions from industrial activities should be 94% below those in 2020, while energy demand can increase moderately by 3% compared to 2020 (IEA, 2021).
Industries in the European Union (EU) are responsible for 9% of global industrial energy demand (13.3 EJ in 2020). To abide by the currently agreed policies, the EU industry will need, in the next 30 years, to limit its energy consumption to 11.7 EJ and in order to meet the climate targets to 9.6 EJ (IEA, 2021). To combat climate change, the EU has put its own climate reduction targets in place. By 2030, GHG emissions should decrease by at least 55% (compared to 1990 levels), the renewable energy share should increase above 32% of the total energy consumption, while energy efficiency should improve by at least 32.5%. In recent plans, an even more optimistic energy efficiency uptake of 36% is proposed (EC, 2020). By 2050, the EU aims to achieve net-zero GHG emissions (EC, 2018a). To achieve the goals, there is a general consensus that a combination of measures will need to be adopted (Åhman et al., 2017;Bataille et al., 2018), from energy efficient technologies to more advanced technologies after 2035, such as Carbon Capture and Storage (CCS), and a switch to CO 2 emission free feedstocks and processes.
Energy efficiency is one of the first steps for climate mitigation (Rosen & Guenther, 2015). This is mainly because energy efficient technologies are readily available and cost-effective, and their adoption tackles several issues from decreasing production costs and increasing the company's comparative advantage to increased energy security and reduced pollution levels (IEA, 2014(IEA, , 2017Rosen & Guenther, 2015).
Detailed case studies estimate the potential for energy savings from the wide implementation of energy efficiency measures across several sectors. Chowdhury et al. (2018) identify the energy savings potential of common energy efficiency measures in the UK iron and steel and the food industries to overcome 15%. Owttrim et al. (2022) demonstrate that energy efficiency alone can reduce CO 2 emissions by 66% in the pulp and paper industry. The non-ferrous metals industry in China can reduce its energy use by 20% through energy efficiency (Shao, 2017) while for certain processes, such as alumina refining, the potential is higher than 30% (Kermeli et al., 2015). The bottom-up analysis of energy efficiency measures identified the economic potential for the Swiss cement industry at 14%, while the technical potentials is around 18% (Zuberi & Patel, 2017). The energy savings potentials identified for the UK chemical industry varied per industrial process from 5 to 35% (Griffin et al., 2018). Bottom-up analyses usually focus on a certain industry or manufacturing process, country or the implementation of specific technology types such as waste heat recovery. Other studies, in a more top-down approach, employ benchmarking practices to estimate the energy saving potentials in the various industrial sectors. Although many longterm energy models have a representation of the industry sector, very few have an holistic technology detailed bottom-up approach, such as in Fleiter et al. (2018), with most not representing the industry 1 Includes the energy use in blast furnaces and coke ovens. The energy use for non-energy purposes, such as feedstock use in the chemical and petrochemical industry (about 36 EJ in 2018), is not included (IEA, 2020). adequately and modelling the energy and GHG mitigation measures in a rather stylized manner (Edelenbosch et al., 2017).
The EU Reference scenario 2016 (Capros et al., 2016) from the PRIMES long-term energy model makes final energy demand projections per industrial sector and EU country, while capturing current policies and market trends. It includes policies and measures adopted in the EU in 2014 and Directive amendments made in 2015, and it assumes that the availability of EU Emissions Trading System (ETS) allowances faces an annual decrease, following current Directive provisions, and industrial energy efficiency is considered to improve as the result of recent policies such as Ecodesign and labelling and the Energy Efficiency Directive (EED). The EU wide GHG reductions from the Effort Sharing Decision (ESD) are also assumed to be achieved in this scenario. The industrial GHG emission intensity slightly decreases (by 2%) in 2020 (compared to 2010) and decreases more drastically in 2030 (27%) and 2050 (51%). This is the result of increased energy efficiency, switch to the production of higher valueadded industrial products, slow growth of energyintensive industries and the shift to less carbon-intensive fuels (Capros et al., 2016).
However, the EU Reference 2016 scenario does not give insights into the extent to which energy efficiency potentials are implemented and thus, to the extent to which energy efficiency potentials remain untapped in such a baseline scenario in the EU. Obtaining these insights will give valuable information on the pace that energy efficiency needs to be incorporated to reach the climate goals and to what industrial sectors each country should invest in energy efficiency or other measures (if energy efficiency is already achieved).
For this reason, this analysis focuses on constructing a Frozen Efficiency scenario that considers the same structural changes as the EU Reference scenario 2016 from PRIMES but with no energy efficiency improvements and recycling uptake. By comparing the final energy demand projections made in these two scenarios, the impact of socio-economic changes and energy efficiency changes on the energy demand projections can be distinguished. In a second step, this study explores four types of alternative future scenarios, able to substantially decrease the final energy demand and/or deeply reduce GHG emissions using detailed bottom-up technological details for the main industrial processes. The developed scenarios have varying degrees of technology diffusion rates and varying types of technological innovations to construct different energy demand pathways for the EU27 + UK industry. All scenarios are based on the Energy Efficiency First Principle (EEFP), meaning that the typical impact on energy use that the use of innovative and fuel switching measures have is adjusted based on the uptake of energy efficiency improvements. For the construction of the different scenarios, no GHG mitigation level is set. The different scenario pathways have been solely developed to identify the technical energy savings potentials under the specific assumptions and to give insights onto the possible impact on energy supply.
The four mitigation scenarios investigated are as follows: • Energy Efficiency scenario: The energy efficiency scenario assumes that BATs are widely adopted across all industrial sub-sectors.

• Energy Efficiency + high recycling scenario:
This scenario has the same assumption as the Energy Efficiency scenario, but it also allows for material recycling improvements in main industries (e.g. increased shares of steel production from scrap The article is structured as follows. The "Method" section gives an overview of how the EU Reference scenario 2016 is broken down and the scenario assumptions for the alternative scenario developments. The "Method" section furthermore includes an overview of the main energy efficiency, innovative and fuel switching technologies per main industrial sector. The "Energy demand projections for the EU industry" section presents the different scenario results for the EU 27 + UK industry, and the "Discussion" section discusses the results and critical assumptions. Finally, the " Conclusions" section presents the main conclusions of the study. The Supplementary Material (SM) includes detailed data on energy efficiency improving and recycling measures, innovative measures and electrification and H 2 measures used in this analysis.

Method
This section describes the process undergone for constructing the different scenarios.
The starting point is the construction of the baseline scenarios. The two baseline scenarios developed are the (i) Reference scenario and the (ii) Frozen Efficiency scenario. The Reference scenario is the same as the EU Reference scenario 2016 from the European Commission (Capros et al., 2016) where current policies are continued but not tightened albeit with more assumptions about sectoral details. The Frozen Efficiency scenario is an alternative baseline scenario that assumes the same socio-economic changes with the Reference scenario, i.e. same changes in industrial value added and production volumes, without however allowing for energy efficiency improvements.
In a next step, the mitigation scenarios are developed. In the Energy Efficiency scenarios, all technologies/measures that could considerably improve the energy efficiency are considered to be widely implemented, and in the two deep decarbonization scenarios, the Electrification and Hydrogen scenarios, the main innovative technologies and the electrification and hydrogen technologies are implemented.
The Frozen Efficiency scenario: breaking down the EU Reference scenario 2016 The EU Reference scenario 2016 (Capros et al., 2016) was the most recent scenario at the time of the analysis with data and projections on an EU country and industrial sector level. The reported industrial energy demand includes the energy used in blast furnaces and coke ovens but excludes feedstocks (e.g. in the petrochemical industries) and primary energy used to produce purchased electricity. Furthermore, refineries are not included.

Method
The base year used in this analysis is 2015, the year for which most disaggregated data on energy consumption were available at the time of the analysis. 2 To construct the Reference and the Frozen Efficiency scenarios, based on the activity levels and base year data used in the EU Reference scenario 2016, the following approach was used (first two steps are made for the Reference scenario and the third for the Frozen Efficiency scenario): (i) The 2015 energy demand reported in the EU Reference scenario 2016 was disaggregated per main industrial sub-sector. The energy demand in the EU Reference scenario 2016 is reported for five industrial sectors, namely iron and steel, pulp and paper, non-ferrous metals, chemicals, non-metallic minerals and others. The industries included in each sector can be very diverse. For example, the non-ferrous metals industrial sector includes the aluminium, copper and zinc sub-sectors. To assess the different energy efficiency options in detail, the energy demand per main industrial sub-sector needed to be identified. This was done by subtracting the product of the typical EU specific energy consumption (SEC) (in GJ/t)  for each main industrial product and the 2015 activity level from the sectoral energy demand reported in the EU Reference scenario 2016. The remaining energy consumption was assigned to the "Rest of…" of each industrial sector. See Eqs. 1-3.
where E Ref ,j,i,2015 , taken from Capros et al. (2016), is the 2015 energy demand in the Reference scenario of industrial sector j in EU country i, and E FrozEff ,j,i,2015 is the energy use in the Frozen Efficiency scenario.
The energy use per industrial product ( E p j ,i,2015 ) and for the "Rest of…" industries of each sector j ( E Rest of j,i,2015 ) was estimated based on Eqs. 2 and 3: (1) E Ref ,j,i,2015= E FrozEff ,j,i,2015 (2) E p j ,i,2015 = p p j ,i * SEC p j 2 A check with more recent statistics from data in 2018 has been performed but energy demand projections were similar. We thereby consider that this analysis for the baseline scenarios can still be representative.
where p p j ,i is the production of industrial product p of the industrial sector j in country i and SEC p j is the specifici energy consumption of product p.
ii) The 2015 energy demand per industrial subsector was broken down per energy carrier. Although the energy carrier breakdown (coal, oil, natural gas, electricity and others) in the EU Reference scenario 2016 is available on a country level, it is not available on an industrial sector level. To disaggregate the industrial energy demand in 2015 per energy carrier (coal, peat, oil, natural gas, electricity, biomass, and waste, geothermal, solar, heat and others) for each industrial sub-sector and country, the IEA (2016a), IEA (2016b) database was used based on Eq. 4.
where c is the energy carrier and FS c,j,i,2015 is its energy share on the overall energy consumption of industry sector j in country i in 2015. The future fuel mix in the Frozen Efficiency scenario was assumed to remain the same as in 2015.
iii) Based on the future activity level from EU (2016), the Frozen Efficiency scenario was constructed. Based on the Reference scenario, a Frozen Efficiency scenario was developed where the SEC remains fixed. The difference between the Reference scenario and the Frozen Efficiency scenario is therefore equal to the (autonomous and policy-induced) energy-efficiency improvement included in the Reference scenario. This provides a good basis for the estimation of the energy efficiency improvement potentials in comparison to the Frozen Efficiency and Reference scenario. To estimate the future energy demand per sector and country in the Frozen Efficiency scenario, the following equations were used: where, where it is assumed that the share of "Rest of…" on the overall energy use of each sector j in each country i ( RatioRest j,i,2015 ) remains fixed to the 2015 level.
Data Table 1 shows the specific energy consumption in the EU for main industrial products. Table 3 shows the production volumes in 2015 from available statistics and the estimated production developments. The industrial activity projections were estimated based on production data from Capros et al. (2016 and other sources (see Table 2). For the industrial products with limited information on the production developments in the EU Reference scenario 2016, value-added assumptions were used. Figure 1 shows the industrial value added 3 per industrial sector (Capros et al., 2016). The main contributor both in 2015 and 2050 is Engineering, responsible for 36% and 45% of total value added, respectively. The most energy-intensive industries, pulp and paper, non-ferrous metals, non-metallic minerals and iron and steel are responsible for 12% of the value added in 2015, much lower than 16% in 1995, and their share is projected to further decrease to 11% by 2050.
Most products experience an increase in production in the 2015-2030 period, while in the 2030-2050 period, most products seem to stabilize (see Table 3). A significant part of the energy-intensive products remains in the EU area (Capros et al., 2016), so there is no expected significant decrease in production. Table 3 shows the production volumes used for the construction of the Reference scenario. The production volumes used in the Reference and the Frozen Efficiency scenarios are the same, except in two cases: i) The clinker produced in the cement industry. In 2015, the average clinker content in the EU was 76% (GNR, 2019). In the EU Reference scenario, it is assumed that the potentials for using recycled materials are exhausted (Capros et al., 2016). We thereby assume that the clinker content in the Reference scenario drops to 66%, which is the lowest clinker content used currently in the EU (ECRA/CSI, 2017). In the Frozen Efficiency scenario, the clinker content remains stable at 76% (current level). ii) The share of steel produced with the electric arc furnace route. In 2015, the share of steel production that uses an electric arc furnace (EAF) was 39% (Worldsteel, 2018). In the Reference scenario, the share of EAF steel is projected to account for more than 42% of total steel production (Mantzos et al., 2020). In the Frozen Effi-ciency scenario, we assume that the EAF share remains stable at 39% in the whole period analysed.

Method
To estimate the energy demand in 2050 in an Energy Efficiency scenario for each of the products ( E EnergyEff ,p j ,i,2050 ), the following equation is used: Table 1 Specific energy consumption (in GJ/t) and energy shares for heating and cooling for main industrial products  a Boulamanti and Moya (2017) where Impl.Rate t,p j ,2050 is the implementation rate of technology/measure t for the industrial product p p j in 2050 and ES t,p j ,2050 is the typical energy savings potential in GJ/t pf product.
In the other mitigation scenarios, Eq. 9 is used to estimate the energy demand ( E DeepMitigation,p j ,i,2050 ). This equation also accounts for energy reductions from implementing energy efficiency measures first.
The ES impr.,t,p j ,2050 is the typical energy savings potentials of a certain technology t (ES t,p j ,2050 ) adjusted for the energy savings implemented from the adoption of energy efficient technologies. ES p j,2050 are the overall savings from previous measures.

Technology descriptions
The following paragraphs describe the key energy efficiency measures, the key innovative measures and the technologies for the Electrification and H 2 scenarios for the six most energy-consuming sectors, i.e.
Impl.Rate t,p j ,2050 * ES impr.,t,p j ,2050 (10) ES impr.,t,p j ,2050 = ES t,p j ,2050 − ES p j,2050 P p j ,2050 cement, chemicals, aluminium, iron and steel, glass and the pulp and paper industries. The full list with all the measures identified and accounted for in this analysis and information on investment costs (in €/t), energy savings potentials (in GJ/t) and current and future diffusion rates are listed in the SM for each of the industrial sectors and main process steps.  (Green, 2007). The two most energy-consuming steps are aluminium smelting (49 GJ/t) and alumina refining (13 GJ/t) (IAI, 2020).

The aluminium industry
Almost all smelters in the EU are of the most energyefficient types, the Point Feed prebake (PFPB) cells, limiting the opportunities for energy reduction from replacing the old and inefficient Söderberg cells (see Table S2 in the SM) (Cusano et al., 2017;Kermeli et al., 2015;Moya et al., 2015). The energy use in EU smelters could be further reduced by improved control systems and improved PFPB cell designs. However, to significantly reduce the industry's energy use, innovative measures, i.e. the use of inert anodes in combination with the wetted cathodes need to be adopted as a retrofit (see Table S8 in the SM). This can reduce the smelting energy use by 3 MWh/t (Moya et al., 2015;Rutten et al., 2017). The alumina refining industry in the EU is small and has not been assessed in this analysis. Reference scenario: cement growth same as in POTENCIA and clinker growth at a slower pace; Frozen Efficiency scenario: cement and clinker grow at the same pace Chemicals EU Reference scenario 2016 (Capros et al., 2016) Fertilizers and inorganic chemicals stabilize and slightly decline in later years, methanol, and ethylene experience strong growth All other industrial products EU Reference scenario 2016 (Capros et al., 2016);  No radical changes Energy use in the secondary aluminium industry, involving facilities where aluminium scrap is remelted, could be reduced with the use of improved waste heat recovery and new de-coating equipment that limits the requirements for scrap pre-treatment (Moya et al., 2015). Foundries are another area that could achieve significant energy savings with the use of improved waste heat recovery (regenerative burners), better process controls and the use of liquid metal as feedstock (Alsema, 2000).
The aluminium smelting industry is already electrified. To electrify the melting furnaces used in secondary aluminium production and casting facilities, induction furnaces could be used to replace the gasfired ones (European Commission, 2005).

The cement industry
The main steps in cement making are (a) raw material preparation, (b) clinker calcination and (c) cement grinding. Most energy efficiency measures focus on clinker calcination, the most energy-intensive step in cement making (3.6 GJ/t) accounting for 90% of total energy use, and most of which is fuel (Worrell et al., 2013).
There are two main ways to decrease the industry's SEC: (1) by retrofitting with energy-efficient technologies/measures such as measures that improve combustion, and waste heat recovery, e.g. the addition of extra preheater stages and of a precalciner, and (2) by clinker substitution with supplementary materials that reduce the need for calcination (see Table S7 in the SM) (Kermeli et al., 2019;Worrell et al., 2013). In the Energy Efficiency + high recycling scenario, it is assumed that cement production relies heavier on clinker substituting materials than in the base year. The assumption has been made that the clinker to cement ratio decreases from 76% in 2015 (GNR, 2019) to 60% in 2050.
The main innovative measure identified is the production of cements with only 25% clinker (Worrell et al., 2013). It is assumed that this measure has an implementation rate of 11% in 2030 and 100% in 2050. Cement making is a high-carbon intensive process. Possible additional deep decarbonization options include the substitution of fuel with biogas, electricity and hydrogen. CCS is another deep decarbonization option for the cement industry but has not been assessed here. Kiln electrification technologies, such as kiln plasma torches and kilns that use microwave energy to provide the heat needed for calcination, are still in the pilot phase (Xavier & Oliveira, 2021). However, it has been concluded that electrification with the plasma burning technique is technically feasible (MPA et al., 2019). Using H 2 to fire cement kilns can be challenging, but some of the combustion issues could be overcome with the use of a biomass-hydrogen mixture (MPA et al., 2019) (see Table S12 in the SM).

The iron and steel industry
This industrial sector is comprised of (i) integrated steel mills that produce steel from iron ore and coke in blast furnaces (BF); (ii) secondary steel mills that produce steel from steel scrap in electric arc furnaces (EAFs), pig and/ or direct reduced iron (DRI); and (iii) iron and steel foundries that melt metal and use moulds to create the final products (Worrell et al., 2010). Steelmaking from raw materials in the so-called primary route is very energy-intensive, requiring in the EU about 18 GJ/t liquid steel (Remus et al., 2012). Energy use for steelmaking from scrap is significantly less energyintensive than steel produced in the primary steel production route.
There is a wide variety of energy efficiency measures/technologies identified in literature for all main process steps such as coal moisture control in coke making, low-temperature heat recovery for cogeneration and improved waste heat recovery (Worrell et al., 2010;Zhang et al., 2019) (see Table S3 in the SM). The sector's energy use can also be reduced through increased recycling, by increasing the volumes of steel produced from scrap instead of raw materials. In 2015, the share of the more energy-efficient steel production route that uses an electric arc furnace (EAF) was 39% (Worldsteel, 2018). In the Reference scenario, the share of EAF steel is projected to account for more than 42% of total steel production (Mantzos et al., 2020). In the Frozen Efficiency scenario, we assume that the EAF share remains stable at 39% in the whole period analysed, while in the Energy Efficiency + high recycling scenario, the share increases to 67% in 2050 (Fleiter et al., 2019). In addition, coke and pig iron production also decrease with the same annual rates as BF/BOF steel.
Innovative or emerging technology measures, such as coke dry quenching and top gas recycling, are implemented only in the Electrification and the Hydrogen scenarios (see Table S9 in the SM). In the Electrification scenario, iron ore electrolysis (Ulcowin, Ulcolysis) is considered to only have a small implementation rate in 2030 (10%), while by 2050, it fully replaces the primary steelmaking route (BF/BOF steel). In the Hydrogen scenario, primary steelmaking in 2050 is replaced by direct iron reduction by H 2 .
The glass industry The glass industry composes a sub-sector of the non-metallic mineral sector. This analysis addressed the two main segments of the EU glass industry, i.e. container glass and flat glass manufacturing. Depending on the glass melting furnace design, the energy consumption can vary from 3 to more than 40 GJ/t (Scalet et al., 2013).
Energy efficiency can be improved through increased levels of cullet and batch preheating, improved burning and advanced controls (Institute for Industrial Productivity (IIP), 2014; Worrell et al., 2008) (see Table S4 in the SM). Fast response programs are only assumed to be widely diffused by 2050 in the Electrification and Hydrogen scenarios. In the Electrification scenario, by 2050, all gas-fired glass melting furnaces are replaced by induction furnaces (see Table S10 in the SM).
The pulp and paper industry This analysis was conducted for the production of chemical pulp, mechanical pulp, recovered fibres, and three types of paper i.e., board and packaging, tissue, and graphic paper. A wide range of energy efficiency measures are available (Fleiter et al., 2012;Harmsen et al., 2018;Rutten et al., 2017) (see Table S5 in the SM), such as improved heat recovery, which is implemented in all scenarios except for the Frozen Efficiency scenario. There are several innovative measures such as black liquor gasification and enzymatic pre-treatment that could further reduce the industry's SEC . In the Electrification scenario, high-temperature heat pumps are assumed to fully replace gas or biomass-fired boilers by 2050 for the heat requirements that are in the range of 100-200 °C. Low-temperature heat pumps are also assumed to fully cover the heat requirements at a temperature lower than 100 °C. In the Hydrogen scenario, all heat requirements are assumed to be covered by H 2 boilers (see Table S11 in the SM).

The chemicals industry
The chemicals industry is a complex industry generating a variety of products. This analysis was performed for the chemical products for which sufficient information could be gathered on energy savings and investment costs of currently available, innovative and fuel switching technologies. These chemicals are ammonia, ethylene, methanol, soda ash and carbon black. No material efficiency measures or recycling were considered for this sector. The energy efficiency measures are widely adopted in all scenarios except for the Frozen Efficiency.
Innovative measures were not identified. Improvements in the compression and separation section with the use of selective membranes were however included in the Energy Efficiency scenario (see Table S6 in the SM). In the Electrification and Hydrogen scenarios, the assumption was made that the conventional processes to produce ammonia, methanol and ethylene are replaced with the low-carbon processes that utilize H 2 as feedstock 4 (see Table S13 in the SM). The adoption of these processes also switches a part, or all the fuel used for energy purposes, to electricity (Bazzanella & Ausfelder, 2017). The conventional processes are generating excess heat (4.3 GJ/tonne ammonia, 1.3 GJ/tonne ethylene and 2.0 GJ/tonne methanol) that in the low-carbon process must be provided otherwise (Bazzanella & Ausfelder, 2017). Here, it was assumed that in the Electrification scenario this heat is provided by electric boilers and in the Hydrogen scenario by H 2 boilers. Another main assumption made is that this heat is required at a temperature higher than 300 °C, the temperature limit for industrial heat pumps. For soda ash production, heat pumps are adopted for the share of the heat needed at less than 500 °C and the rest using electric or hydrogen boilers. For the manufacturing of carbon black, we have not included technologies for fuel switching due to the limited data availability.

Scenario assumptions
The summary of all assumptions per scenario and industrial sub-sector is listed in Table 4.
To construct the different scenarios, the implementation of several technologies/measures was assessed, clustered as follows:  ii) Material efficiency measures. These are measures that lead to a reduction in the amount of primary material needed to provide a specific material service (Worrell et al., 1995). For example, using more steel scrap in steel manufacturing will reduce the need for iron ore, and using more clinker substituting materials in cement making will reduce the need for limestone . iii) Innovative measures. These are technologies that can offer significant energy and GHG reductions compared to conventional technologies but that are currently not available for wide implementation due to the low TRL. They include both energy efficiency measures and material efficiency measures. iv) Electrification measures. These measures refer to fuel-switching technologies that can replace fossil-fuel based technologies with technologies that use electricity. v) Hydrogen measures. These measures refer to fuelswitching technologies that can replace fossil-fuel based technologies with technologies that use hydrogen.
Although material efficiency measures and some of the electrification and hydrogen measures can decrease the energy intensity of the process, they are not included under the group of the energy efficiency measures. The adoption of the different measures for each scenario is based on exogenous assumptions.
In general, first production decreases based on the assumed material efficiency rates, 5 then all energy efficiency measures in the form of BATs are adopted, then all innovative measures and lastly all fuel switching measures. In this way, first the reduction of the industrial energy demand is quantified and then the effect of fuel switching on the energy demand.

Energy demand projections for the EU industry
This section first presents the Frozen Efficiency and Reference scenarios in terms of projected final energy demand towards 2050. Then, the developed mitigation scenarios are included for further comparisons.

Frozen Efficiency and Reference scenarios
According to the Reference scenario, similar to the EU Reference scenario 2016 (Capros et al., 2016), the industrial energy demand in the EU is shown to decrease from 11.9 EJ in 2015 to 10.6 EJ in 2050. After a short increase in the first 5 years, it decreases annually by 1% in the 2020-2035 period and by 0.1% in the 2030-2050 period (see Fig. 2). This is the result of (i) energy efficiency improvements and (ii) structural changes in the industrial activities which are assumed to move towards less energy-intensive and higher value-added products (Capros et al., 2016).
The constructed Frozen Efficiency scenario includes the same industrial structural changes as the Reference scenario but without any energy efficiency improvements. The industrial energy demand, without energy efficiency improvements, is found to increase to 14.6 EJ by 2050, at an annual growth rate of 0.6%. The increase is more prominent in the 2015-2030 period where the production growth is stronger (see Table 3).
In 2015, five countries, Germany, France, Italy, UK and Spain were responsible for 59% of the total industrial energy demand in the EU. The same countries are still projected to account for most of the industrial energy use in 2050 (57%) in the Reference scenario, while in the Frozen Efficiency scenario, the share remains the same (59%). Figure 3 shows the developments of the main industrial sectors in the EU27 + UK in the period 2015-2050 in the two scenarios. In the Reference scenario (0.1% annual decrease in 2015-2050), the sectors that decrease their energy demand are chemicals (25%), iron and steel (14%), paper and pulp (29%), non-ferrous metals (17%) and non-metallic minerals (15%). The Others sector is the only sector increasing its energy demand by about 6%. In the Frozen Efficiency scenario, all sectors increase their final energy demand: chemicals (8%), iron and steel (9%), paper and pulp (10%), non-ferrous metals (8%), non-metallic minerals (21%) and Others (41%).
The Frozen Efficiency scenario allows for certain structural changes, i.e. the switch to higher value-added products, but it does not allow for (i) higher rates of the EAF route and (ii) higher clinker to cement ratios than in the base year. Figure 4 shows the industrial energy demand in the EU industry when such changes are on the same level as the Reference scenario. When these changes are allowed, the energy demand in the Frozen Efficiency scenario increases to 14.1 EJ in 2050 instead of 14.6 TJ when these changes are not included.
Increasing the EAF share from 39% (EU 2& + UK average in 2015) to 43% will reduce the energy demand by approximately 140 PJ in the iron and steel industry. Decreasing the clinker to cement ratio from about 76% (EU 27 + UK average in 2015) to 66% will reduce the 2050 energy demand in the non-metallic minerals sector by about 160 PJ (see Fig. 4). Table 5 shows the annual autonomous and policyinduced energy efficiency improvements compared to the base year (2015). It is calculated by annualizing the difference in the final energy demand (fuel or electricity) between the Reference scenario and the Frozen Efficiency scenario for each industrial sub-sector. The highest fuel efficiency improvements in the 2015-2050 period are observed in the pulp and paper, non-ferrous metals and chemicals sectors. It is also observed that the highest rates of improvement are in the period 2020-2035 ranging from 0.0 to 2.1%. The improvements are lower in the case of electricity, but still, the same trend is observed, i.e. the improvement is stronger in the 2015-2035 periods relative to the later period. Figure 5 shows how the different energy carriers develop in the two scenarios (Reference and Frozen Efficiency) during the 2015-2050 period. In the Reference scenario, the share of coal products on the overall energy use decreases from 15% in 2015 to 9% in 2050, for natural gas from 29 to 22% and for oil from 10 to 6%. The shares of electricity, biofuels and heat increase from 30%, 9% and 6% to 39%, 15% and 9%, respectively. Since in the Frozen Efficiency scenario the shares of the different energy carriers remain stable per sector throughout the analyzed period, the energy mix in 2050 is different from the Reference scenario and does not change much compared to the base year 2015. In 2050, coal accounts for 14%, natural gas for 30%, oil for 11%, biofuels for 9% and electricity for 30%, in the Frozen Efficiency scenario. The shares of biofuel and heat remain at the 2015 levels.

Mitigation scenarios
It is found that in the Energy Efficiency scenario, the wide adoption of energy efficiency improvements can reduce the industrial energy demand in 2050 from 14.6 EJ in the Frozen Efficiency scenario to 12.6 EJ, an energy savings potential of about 14%. About 200 PJ can be saved in the chemicals industry, 360 PJ in the iron and steel industry, 380 PJ in the non-metallic minerals industry, 110 PJ in the non-ferrous metals industry, 75 PJ in the pulp and paper industry and around 800 PJ in the Others industry. When increased recycling or material efficiency is also considered for three industries (iron and steel, cement and aluminium), the 2050 final energy demand further reduces to 11.3 EJ. This is an energy-saving potential of 23%, compared to the Frozen Efficiency scenario.
Comparing the Energy Efficiency + high recycling scenario to the 27% energy savings obtained in the Reference scenario, where energy demand decreases due to BAT implementation and only incremental recycling is allowed, it is evident that the projected energy savings in the Reference scenario are very optimistic. In the Reference scenario, final energy demand in industry amounts to 10.6 EJ in 2050, which is lower than the final energy demand in the BAT (high recycling) scenario where energy demand reduces to 11.3 EJ. To reach the 2050 final energy demand in the Reference scenario, in addition to the wide adoption of BATs, and high recycling levels, more measures such as increased material efficiency and innovative measures need to be implemented.
The innovative measures identified in this analysis have the potential to decrease the final energy demand by at least another 500 PJ. In the Electrification scenario, the final energy demand was calculated to decrease to about 9.7 EJ, an energy savings potential compared to the Energy Efficiency + High Recycling scenario of approximately 14%. In this scenario, about 73% of the energy demand is covered by electricity and the rest by fuel consumption. In the Hydrogen scenario, the final energy demand was found to reach 10.3 EJ, 7% higher than the electrification scenario. This is because industrial heat pumps that operate largely on waste heat are not included in the Hydrogen scenario.
In the Hydrogen scenario, 30% of the final energy demand is covered by fuel consumption, 42% by electricity and 29% by H 2 .
Vol.: (0123456789) The energy demand presented in Fig. 6 is final energy demand, i.e. the conversion losses for the production of electricity and H 2 are not included. The degree to which fossil energy savings occur depends on the energy carrier mix used to generate the electricity and H 2 . Figure 7 shows the 2050 energy demand under the electrification and the H 2 scenarios in the cases where (i) material efficiency (incl. recycling) is not included, (ii) material efficiency and innovative measures are not included, and (iii) material efficiency, innovative measures and energy efficiency measures are not included. It can be seen that without the uptake of energy reduction measures the energy demand can be 27% higher in the H 2 scenario and 37% higher in the electrification scenario. In such a scenario, the H 2 demand is 56% higher and the electricity demand 16% higher in the H 2 scenario. Also, in the electrification scenario, the electricity demand would be 30% higher. Not capturing the full energy savings potential that current and future energy efficiency technologies can offer will result in larger energy demand and significant system energy losses from the production of electricity and H 2 that could be avoided.

Discussion
The scenarios analyzed present possible future developments while using the industrial activity assumptions on the socio-economic development taken by the EU Reference scenario 2016 (Capros et al., 2016). In addition to the assumptions made for future industrial activities, several sets of assumptions have been made.
The analysis covers the whole EU industrial sector. However, not all industrial sub-sectors have been analyzed in detail. For the "Rest of.." sub-sectors, such as the bricks and clay industries that constitute a part of the non-metallic minerals sector and the copper and zinc industries that are part of the non-ferrous metals sector, the savings are instead assumed to be the average of the sector they belong to. In addition, for the Others sector, the savings are an extrapolation made based on the total savings of the sectors for which the detailed analysis was performed. This however might be an underestimation of the energy savings since the industries analyzed are the most energy-intensive, with high energy costs and might already be quite efficient.
For these two industrial segments, i.e. the Others sector and the "Rest of …" sub-sectors, in the Electrification and the Hydrogen scenarios, the savings/ energy demand was estimated from the application of electric and H 2 boilers, for the share of the energy demand used to provide heat at temperatures below 500 °C (see Table S1 in the SM). Due to the limited information available, no other fuel switching measures were considered. The potentials thereby in the Others and the "Rest of …" industries can be higher as also other technologies (e.g. electric furnaces) can be implemented. This analysis accounted for electrification and H 2 measures for approximately 82% of total final energy demand.
Another main assumption concerns the implementation rates for 2030 and 2050 and the diffusion rates of the different technologies in 2015. Although Fig. 4 Final energy demand in the EU27 + UK iron and steel and non-metallic minerals industry in the Reference and the Frozen Efficiency scenarios the 2015 diffusion rates differ per measure, due to the lack of data, we assumed that they are the same for all countries. The same applies to the average energy intensities of the various products manufactured that were assumed to be the same for all EU27 + UK countries. However, because the energy demand for the manufacture of intermediate products (e.g. clinker used for cement making, coke used in primary steel making, steel from scrap and steel from pig iron) was also investigated, the specific energy intensities for certain final products differ per country. Although this is an issue for country estimates, it is not for the estimates on an EU level (average EU SEC drops to the BAT or new technology level).
This bottom-up industry analysis was based on the EU Reference 2016 scenario (Capros et al., 2016). A In the latest EU Reference 2020 scenario (Capros et al., 2021), the 2050 industrial energy demand in the EU27 is projected to be 10% lower than in the EU Reference 2016 scenario. The share of natural gas on the overall energy use is slightly higher (27% compared to 24%), while the use of solid fuels is the same (around 9%). The production developments in the Frozen Efficiency scenario, used as the baseline scenario for all mitigation scenarios, were based on the EU Reference scenario 2016. Different production trends would thereby have important impact on the results of this analysis. Overall, the assumptions on production activities made in the EU Reference scenario 2020 for the long term are similar to the 2016 scenario, i.e. the switch to the less energy-intensive industries, the maintained activity in the iron and steel and the nonferrous metals industry, the increase in the share of engineering on the overall energy use and the stagnation and slight decline of the fertilizers and inorganic chemicals. No information was available on the methanol and ethylene production. Since the projections available in the EU Reference 2020 scenario are not available per industrial sector, a more detailed comparison is hard to perform.

Conclusions
According to the EU Reference scenario developed by Capros et al. (2016), the industrial energy demand in the EU 27 + UK is expected to slowly decrease from 11.9 EJ in 2015 to reach 10.6 EJ by 2050, with an annual decrease rate of about 0.3%. Although the energy demand is considered to rise in the EU due to the increased level of industrial activities, the uptake of energy efficiency improvements in combination along with the switch of activity from the energyintensive industries to less energy-intensive industries is found to outweigh the increasing energy demand trends.
In a Frozen Efficiency scenario, the energy demand in 2050 will be 22% higher compared to 2015, reaching about 14.1 EJ. The Frozen Efficiency scenario is a variation of the Reference scenario that although it considers changes in production developments, no energy efficiency improvements are included. In the Frozen Efficiency scenario, industrial energy demand increases at an annual rate of 0.6%. The increase is more prominent in the 2015-2030 period where the production growth is stronger. The energy demand grows stronger in the others (41%) and the non-metallic minerals (21%) sectors, followed by the pulp and paper (10%), iron and steel (9%), chemicals (8%) and the non-ferrous metals (8%) sectors.
The savings already included in the Reference scenario reach 27% of final energy demand in 2050. Disaggregated per industrial sector this amounts to 35% for the pulp and paper industry, 32% for the chemicals industry, 30% for the non-metallic minerals industry, 24% for the non-ferrous metals industry and 20% for the iron and steel industry. Comparing the Reference scenario to the constructed Frozen Efficiency scenario allowed for the identification of the energy savings already included in the EU Reference scenario 2016.
Energy efficiency improvements alone can decrease the industrial energy demand by 13.5% in 2050. The wide adoption of energy efficiency technologies/measures can reduce the 2050 final energy demand from 14.6 EJ in the Frozen Efficiency scenario to 12.6 EJ. Of which, about 200 PJ can be saved in the chemicals industry, 360 PJ in the iron and steel industry and 380 PJ in the non-metallic minerals industry, 110 PJ in the non-ferrous metals industry, 75 PJ in the pulp and paper industry and around 800 PJ in the Others industry.
Energy efficiency improvements in combination with increased levels of recycling can decrease industrial fuel use by 27% and electricity use by 11%. When   Fig. 7 Final energy demand in the Electrification scenario (left) and the H 2 scenario (right) without the inclusion of measures/technologies that decrease the energy demand increased recycling or material efficiency for three industries, i.e. iron and steel, cement and aluminium are considered, the 2050 final energy demand further reduces to 11.3 EJ. This is an energy-saving potential of 23% compared to the Frozen Efficiency scenario.
Energy efficiency improvements alone cannot reach the savings estimated by PRIMES in the EU Reference scenario 2016. When analysing the above scenarios, the 27% energy savings included in the Reference scenario cannot be reached only with increased energy efficiency and an incremental increase in recycling rates. To reach the 2050 final energy demand in the Reference scenario, in addition to the wide adoption of energy efficiency measures, high recycling levels, increased material efficiency and innovative measures will need to be implemented.
To decrease the energy demand even further, additional measures are needed. The innovative measures identified in this analysis have the potential to decrease the final energy demand by at least another 500 PJ. In an Electrification scenario, the final energy demand in 2050 decreases by 34% and in a Hydrogen scenario by 29% when compared to the Frozen Efficiency scenario. To achieve these savings, electricity and H 2 will need to be generated by mainly renewable sources. Future research needs to address how the energy efficiency technologies and measures introduced in this study can be combined, and supplemented with further options, to enable a transition to a 100% renewable energy supply in industry. In addition, this assessment could be complemented with the flexibility that the industry sectors could offer, such as with demand response management measures, in an 100% renewable energy system. Energy and material efficiency measures limit considerably the demand for energy. Without the wide adoption of energy efficiency, material efficiency and innovative energy efficiency measures, the 2050 industrial energy demand is estimated to be 37% and 27% higher, in an Electrification and a Hydrogen scenario, respectively, compared to when these measures are adopted. Energy efficiency has thereby an important role to play when considering the system integration and optimization for the industrial energy transition.
Funding This paper is part of the sEEnergies project which has received funding from the European Union's Horizon 2020 Research and Innovation Programme under Grant Agreement No 846463 (www. seene rgies proje ct. eu).

Conflict of interest The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.