Life cycle assessment and shadow cost of steam produced by an industrial-sized high-temperature heat pump

Heat pumps are considered to be a promising technology to mitigate industrial emissions from steam provision. In this work, an existing scale-up framework for generating life-cycle inventory data for residential heat pumps was tested if applicable to large-scale high-temperature heat pumps. The benchmark was experimental data from existing machinery. It was found that the scaling framework underestimated the total mass; thus, a new scaling framework was developed and was applied to data obtained from a test run of a commercially available high-temperature pump. A full life cycle assessment was performed by comparing two different steam pressure levels, 2 and 5-bar, produced by the high-temperature heat pump to relevant benchmarks, such as steam from a fossil fuel-driven steam boiler. Although greenhouse gas emissions were reduced by as much as 98% in the best-case scenario, other midpoint categories exhibited a more mixed result in which sometimes the heat pumps were favorable and at other times the benchmarks had a lower impact. An unexpected finding was that the working fluid and its leakage did not have a significant contribution to the global warming potential but it was almost solely responsible for the ozone depletion potential. Recommendations were derived on how to further improve the environmental impact of high-temperature heat pumps. Ultimately, the life cycle impact assessment results were converted into shadow price to allow for a comparison between different areas of protection and to give policymakers an estimate of the total externalized cost of the technology.


Introduction
Global efforts are necessary to lower CO 2 emissions and reduce the effects of climate change.In the European Union (EU), the planned path to reach its climate goals are laid out in the European Green Deal (EU Economy and Society to Meet Climate Ambitions, n.d.).The outline of these climate goals is reducing CO 2 emissions by 55% by 2030, compared to 1990 levels, and becoming the first carbon-neutral continent by 2050.These targets will also affect the industrial sector by means such as the EU emissions trading system (European Commission.Directorate General for Communication, 2021) which is challenging, as the industrial sector is known to be particularly energy intensive.According to the International Energy Agency (IEA), in 2019, the world's total primary energy supply (TPES) was 168,519 TWh.Out of this, only 55,680 TWh (approximately 33%) was used for industrial processes and the rest was wasted, mainly in the form of heat dissipated to the environment (International Energy Agency, 2021).A more detailed analysis on the temperature shares and heat demand by heat levels of various industries were elaborated in Puschnigg et al. (2021) As the majority of the primary energy demand worldwide is satisfied with fossil fuels, a higher energy demand automatically leads to increased greenhouse gas emissions (Forman et al., 2016).
The high temperature heat pump (HTHP) examined in this study was designed for implementation in the pickling line of a steel plant in Spain.This sector accounts for 8% of the global final energy demand and emitted 3.4 Gt of CO 2 in 2019 (International Energy Agency, 2020).The primary steel-making process is responsible for the largest part of these emissions, but a significant amount of CO 2 is also emitted at the finishing line.A recent study identified 1% of the total energy demand for Austrian steel making in 2020 goes to the finishing line, which is mostly supplied by natural gas or off-gases (Beck et al., 2023).What seems like a small number quickly is put into perspective upon the fact that every steel plant needs this kind of energy.Heat pumps show great potential because of the lower temperatures in these processes than in the primary production steps.Especially steam is an important energy carrier in these applications.Heat pumps can decarbonize these processes through electrification.Although the steam-generating heat pump considered in this study was designed to be implemented in the finishing line of a steel plant, the results of this study can be applied to many other industrial sectors, as steam is a universal heat carrier for many industrial processes.This heat pump could be used in any other process that requires lowpressure steam of up to 5 bar, as discussed later in the introduction.Steam-generating heat pumps are relatively new to the market but offer great potential to replace fossil-fuel-driven steam boilers by making use of waste heat from other industrial processes.An overview of the commercial products and demonstrators of research projects is provided in Task 1 of Annex 58 by the International Energy Agency (IEA) (Task 1-Technologies, n.d).
However, one must not forget that sustainability means more than reducing greenhouse gas emissions.Protection of habitats as well as preserving clean air, soils, and bodies of water are of equal importance.
Life Cycle Assessments (LCA) have become one of the most renowned tools for assessing the sustainability and environmental impacts of these parameters for countless chemicals, products, and services.Heat pumps (HP) are no exception to this circumstance, as there is abundant literature on LCAs of residential heating provided by HPs (Lozano Miralles et al., 2020;Ristimäki et al., 2013;Shah et al., 2008).The literature on High-Temperature Heat Pumps (HTHPs) and steam-generating heat pumps is substantial.Note that in this study, HTHP and steamgenerating heat pumps are used interchangeably.The literature is more focused on the practical research and development of HTHP systems, e.g., Frate et al., investigated the proper selection of working fluids depending on heat sink and source temperature while Abedini et al. investigated the question whether mixtures of working fluids provide an advantage (Abedini et al., 2023;Frate et al., 2019).Liu et al., wrote about new pathways to utilize waste gases (Liu et al., 2022).In a recent work, Dai et al., simulated HTHPs and calculated greenhouse gas emissions with emission factors and found a 58% reduction when compared to coal-fired boilers (Dai et al., 2023).There are also works dealing with the sustainability and policy aspects of the technology.An analysis of the heat potential of HTHPs in the EU-28 by Marina et al. highlighted a potential energy output of 23 GW or 641 PJ (Marina et al., 2021).Arpagaus et al., found a similar value of 626 PJ of heat potential under 150 • C (Arpagaus et al., 2018).Similar analysis of potentials are also found for individual countries, as Obrist et al., identified a potential for HTHPs in Switzerland of 1 GW by 2050, while Dumont et al.,found 12 TWh of heating demand in German food industry which could be substituted with steam from HTHPs (Dumont et al., 2023;Obrist et al., 2023).In the work of Arpagaus et al., the Ozone Depletion Potential (ODP) and Global Warming Potential (GWP) of working fluids were discussed, yet studies on full LCAs on HTHPs remain scarce.One major hurdle in this endeavor is that no inventory data for HTHPs are found in the LCA databases and literature.Jiang et al., found in a recent review that the main areas for improvement are diverse when it comes to HTHPs.While there are industrial-sized systems that are either using low-GWP working fluids, have high temperature lifts or COPs >4 the authors concluded that the ideal system in the future should have all of these features (Jiang et al., 2022).If this demand cannot be filled, tradeoffs are poised to be weighed against each other with systemic analysis such as LCA.
Scale-up and future developments often pose challenges in assessing novel technologies and complex mathematical models.For smaller-scale residential HPs, Caduff et al. developed a model that correlates parameters such as the working fluid (WF) use and Coefficient of Performance (COP) with the HP mass based on the empirically found power law often used when scaling up machinery (Caduff et al., 2014).However, no such framework exists for larger industrial applications.
To close this research gap in LCAs for HTHPs, the model from Caduff et al. was benchmarked against commercially available HTHP systems in a forthcoming conference paper by our group (Zeilerbauer et al., 2023).Moreover, the scaling and modelling of working fluids were also examined (Zeilerbauer et al., 2023).The results obtained in these theoretical cases are used to model a cradle-to-gate LCA system for the previously mentioned HTHP system, which was tested on-site and firsthand data was used to assess the machine.The functional unit was 1 GJ of steam produced which was benchmarked to common other steamgeneration technologies.
In order to facilitate decision-making, the shadow price for environmental damage was calculated to give researchers, policymakers, and industry executives a more rounded and balanced view on the matter that aids in helping the decision-making process whether to invest in HTHP technology.The issue with LCA methods is that they usually consist of >10 damage categories and hence different results dealing with different areas, such as climate change, toxicity, land-use, or ozone depletion are obtained.Naturally, it is difficult to find weighting factors between these very different results; however, experts have proposed using monetary values early-on (Guinée et al., 2002).One of those approaches is the before-mentioned shadow price, which has the underlying idea that in the end a government, representing its people, has to "repair" environmental damage and is therefore interested in mitigating these costs altogether as discussed by Hadjimichael et al. (2016) based upon (Wu et al., 2005).

Materials and methods
In this section the procedures for obtaining the data needed, both through work in the laboratory and the subsequent calculations, are presented.

BAMBOO high temperature heat pump system
The investigated machinery is presented in Fig. 1.The HTHP used was a Viking HeatBooster HBS 4 system, which has a nominal heat supply capacity of 250 kW (Viking Heat Engines, 2019).Additionally, the system featured an additional, custom-designed flash tank for steam generation, as the Viking HeatBooster HBS 4 system is not able to produce steam when run as a stand-alone system.This combination of machinery is from this point on referred to as "BAMBOO HTHP" for the remainder of this work, named after the funding project (European Commission, 2023).The setup is composed of a water-to-water hightemperature heat pump and a flash tank unit.The flash tank unit evaporates the pressurized hot water from the heat pump which can then be drawn from the system as saturated steam.This system was completed by a circulation pump which was used to pump the compressed hot water to the flash-tank for the subsequent evaporation.A more schematic overview is provided in Fig. 2.

Life cycle assessment
The general LCA method has been described in the literature before in great detail.Therefore, we would like to direct interested readers to the relevant standard works (Frischknecht, 2020;Guinee, 2002;Klöpffer and Grahl, 2014).This LCA was conducted in the four phases introduced in ISO 14044.
1. Definition of Goal and Scope 2. Life Cycle Inventory (LCI) analysis 3. Life Cycle Impact Assessment (LCIA) 4. Interpretation and Discussion

Definition of goal and scope
The goal of this study was to assess the steam produced by the BAMBOO HTHP system.The system boundary was cradle-to-gate, as no assumptions about steam usage were made, with a functional unit of 1 GJ of steam produced.The 1 GJ of steam refers to the amount of heat exchanged within the condensing unit.The system entailed the production of machinery and usage phase, which was assumed to be 15 years.The machines were assumed to operate 8000 h per year, and the end-of-life operations were not modelled in this study.The plant's production featured mostly European data and German electricity mix for manufacturing, as the company was based in Germany.Ultimately, the environmental impacts caused by 1 GJ of steam by the BAMBOO HTHP were compared with 1 GJ of steam from other common sources.  .Schematic overview of the high temperature heat pump system.In contrast to a regular water-to-water high-temperature heat pump this setup employs an extra circuit composed of circulation pump, flash valve and flash tank to generate steam.A portion of the pressurized hot water flow, driven by the circulation pump, is expanded and evaporated at the flash valve.Steam can then be drawn from the flash tank while the unevaporated part of the flow is recirculated.

Life cycle inventory
The LCI analysis was two-fold.The first part dealt with the production of the HTHP and the materials and power required for construction.This was the result of the ecoinvent processes "Heat Pump Production, brine-water, 10 kW" scaled up according to the findings of our scale-up approach for HTHPs mentioned in the introduction and accepted for presentation at IEA's 14th Heat Pump Conference (Zeilerbauer et al., 2023).
For the BAMBOO HTHP, data were obtained using a test run on-site in France.The exact experimental setup and calculation of relevant parameters for this study were first presented at the DKV Conference 2022 in Magdeburg, Germany (Riedl et al., 2022).

Life cycle impact assessment
For the characterization of environmental damages, CML 2001-2016 LCIA method was used (Guinée et al., 2002).GaBi 10.6 software was used for modelling, as well as for the calculation of results ('Life Cycle Assessment (LCA) Software ', n.d.).Data sets were sourced from Sphera and Ecoinvent and the respective datasets are described in greater detail later in this work.

Sensitivity analysis
Ultimately, after calculating the first set of results, a sensitivity analysis was performed.Therefore, different parameters which were found important for certain impact categories were altered in certain ranges.Thus, the relative change could be monitored and discussed.

Obtaining data for BAMBOO high temperature heat pump test case
All the following equations were adapted from the conference talk and paper by Riedl et al. (2022).The coefficient of performance (COP) is one of the most important parameters in HPs and HTHPs.The system at hand was derived according to Eqs. ( 1) and (2).The power consumption P of both pumps was measured.(2) To determine the heat supply capacity, the QHTHP , Eq. (3) was used.The volume flow in the circulation, Vc , and the temperature difference reached within the heat sink, ΔT Heat Pump,Sink , were measured directly.The temperature dependent properties, such as the density of water, ρ Outflow Heat Sink , and specific heat capacity, c p Outflow Heat Sink , were obtained using the opensource library CoolProp (Bell et al., 2014).
Ultimately, 21 different points of operation were evaluated.In this study 3 different points were selected.An overview of this process is shown in Fig. 3.The three points selected represent ideal cases within their respective pressure range, as high COPs are reached.This selection is discussed in greater detail later in this work.

Scale-up for the LCI
As mentioned in the introduction, the scaling approach of Caduff et al. was evaluated.In the original work (Caduff et al., 2014), the following scaling law was mentioned, which has been used and developed further in numerous other research works (Jalili et al., 2023;Mantei et al., 2022;Peters et al., 2003), as shown in Eq. ( 4).
where C is the cost/material requirement of interest, a is the normalization factor, X is the capacity/scale of interest, and b is the scaling factor.If this factor b is 1, the relationship is linear; however, data showed a value of 0.6, which is best suited for many different applications (Peters et al., 2003).In the work of Caduff et al., an effort was made to correlate the HPs mass with COP and WF (Caduff et al., 2014).Thus, the general form of Eq. ( 4) is transformed logarithmically, as shown in Eq. (5).In this case, i is the property of interest, overall mass, COP, or mass of working fluid, and a and b are scaling factors.The scaling factors by the original authors (Caduff et al., 2014) were compared to data from commercially available HTHPs data compiled in the IEA Annex 58 of Task 1 (Task 1-Technologies, n.d) in our conference talk (Zeilerbauer et al., 2023).The values found by Caduff et al. and used in this work can be found in the Supporting Information (Caduff et al., 2014).

Evaluation of working fluid impacts
Modelling WFs in HTHPs poses a challenge in LCA.Only R134a (1,1,1,2-Tetrafluoroethane), which is one of the oldest and most common refrigerants, can be found in the Ecoinvent 3.8 database.This may be due to a wide variety of (complex) synthesis routes and comparatively little interest in the LCA community.For the case at hand, R1336mzz(Z) ((Z)-1,1,1,4,4,4-hexafluoro-2-butene), Zhang et al., highlighted 6 different precursors suitable for conversion to R1336mzz(Z), with more routes to obtain said precursors (Zhang et al., 2016).Another issue is the lack of LCIA data in literature.For many WFs, including R1336mzz(Z), only the ODP and GWP values have been disclosed (Dai et al., 2015(Dai et al., , 2020;;Zhang et al., 2016).Therefore, we decided to use R134a as a proxy in this study.While it is recognized in the literature (Canals et al., 2011;Moni et al., 2020) that this approach induces uncertainty, LCIA results over all midpoint categories can be obtained.

Test-case data for the high temperature heat pump system
Data obtained for the three operating pressures are found in Table 2.The heat supply capacity was calculated according to Eq. ( 3).This is the heat transferred in the heat sink of the HTHP for steam-generating purposes.Another entity of interest is the heat content of the steam, which is found by multiplying the mass of steam with the latent heat.Since in this case no additional heat is added or extracted (except for heat losses in the flash tank system) these two entities coincide within their measurement uncertainty, as shown in Riedl et al. (2022).
The considered operating points do not reach the full thermal supply capacity of 250 kW of the heat pump.This is not unusual since it only reflects the maximum heat supply capacity of the heat pump and highly depends on the operating point.Interestingly, the results between 3.5 and 5 bar are very similar therefore, within this pressure range, the additional amount of steam is balanced.Additionally, the performance of the heat pump and flash tank setup shows room for improvement, since it is only in its early design phase and more time in testing is needed to give a complete representation of the performance of the setup.This is also the reason why the 3.5 bar conditions show no advantage over the 5 bar conditions.One limiting factor of the performance of the overall setup is the overpowered circulation pump, which was needed to drive the water flow through the flash valve where a part of the pressurized hot water evaporates to become steam.In the design phase, it was chosen to be applied in a larger operating range than needed.This results in inefficient operation in the testing conditions at hand.A more detailed discussion on the performance of the setup can be found in Riedl et al. (2022).

Energy demand and output
As indicated in the previous section, the 3.5-bar operating point was omitted because of its similarity to the 5-bar case.This was because of the approach of comparing only the final energy output.The authors are well aware that not all steam pressure levels are suitable for all applications, yet this simplification was used to make the LCA easier in terms of comprehension.The energy LCI was derived assuming a lifetime of 15 years for the machinery and an annual operating time of 8000 h.Therefore, the total energy output and additional data pertaining to the steam produced are presented in Table 3.A different HTHP prototype investigated by Mateu-Royo et al. yielded a highly similar COP for similar parameters (120 • C heat sink out) (Mateu-Royo et al., 2019b), thus confirming the result obtained.
In the LCA model, two electricity sources were examined for powering the HTHP.The following processes from Sphera's GaBi Professional database were used to examine the environmental effects of immediate installation, without waiting for further decarbonization of the current grid.Wind power was used to answer the question of whether a fully decarbonized power source is necessary to be preferential from an environmental perspective.The steam produced by these two power sources was benchmarked against four commercially available solutions.Two of them were biobased and two were fossil-based, all with a Spanish geographical scope.

Material balance
In order to derive a material balance, a number of assumptions and scaling approaches had to be undertaken.In our previous work (Zeilerbauer et al., 2023), the scaling approach shown in Eqs. ( 4) and ( 5) of residential HPs by Caduff et al. (2014) did not transfer to HTHPs when benchmarked with real HTHP data.Details and documentation for the singular data points and the HTHPs examined can be found in the Supporting Information.The same goes for the solutions for Eq. ( 5), original found by Caduff et al. (2014).The technical documentation by the manufacturer states that the Viking HTHP's mass is 3000 kg for the main unit and electrics, but does not offer more detailed information.This data point is presented within the scale-up framework in Fig. 4. Similarly to the data compiled by the IEA Annex 58 of Task 1 (Task 1-Technologies, n.d) of commercially available HTHPs, the BAMBOO HTHP confirms the trend of a linear scale-up of the 10 kW Ecoinvent process (Heck, 2010) showing the most accurate fit.For the exact mass of the LCI, the scaling/fitting results of (Zeilerbauer et al., 2023) were also amended with the BAMBOO HTHP, as shown in Fig. 5.
When comparing the linear fit of all data points and those <700 kW, no considerable difference is visible at first, as the BAMBOO HTHP lies close to the intersection of both fits.If one applies the fit from the lower (< 700 kW) data points, the equation yields a total mass of 3031 ± 538.33 kg for the BAMBOO HTHP-The fit for all the data points was 2906 ± 3530 kg for the BAMBOO HTHP.Linearly scaling the 10 kW Ecoinvent process (Heck, 2010), which amounted to a material demand of 128 kg for 10 kW, yields 3200 kg.In conclusion, with a standard deviation of 121%, the fit over all data points can be disregarded as too imprecise.This is not surprising, with data ranging from 30 to 1000 kW.Although both the lower and linear approaches overestimate the material balance, the error seems acceptable.In this case, the linear approach overestimated the material demand, as found in the original work by Caduff et al.However, the slopes they found, as shown in Fig. 4, would have underestimated the HTHPs mass (Caduff et al., 2014).It can be concluded that there is no single scaling framework applicable to both HPs and HTHPs.
For further LCA, the 3031 kg found by the lower data fit were used, and the uncertainty was further examined and discussed in the sensitivity analysis of the work.While the manufacturer's documentation outlined 3000 kg of mass, we decided that the difference was negligible and stayed with the linear scale up-approach.The material balance of the original Ecoinvent process, which was scaled linearly in order to build the material balance for the machinery of the BAMBOO HTHP and the linearly scaled values to 3031 kg are listed in Table 4. Ecoinvent and Sphera data were used for modelling the processes.Sphera processes are featured with a direct quote to their respective documentation page, whereas Ecoinvent processes are named and the database version number is listed after the semicolon (Wernet et al., 2016).
Concerning the WF 134a, the documentation of the original process mentioned a 20% loss after filling and during the deconstruction of the machine (Heck, 2010).After consulting HP experts, this value was deemed too high.Instead, an initial filling of 65 kg was chosen, with annual leakage of 1%, which led to a total demand of 74.75 kg, when refilled.The leakage was modelled as an emission to air.
The scale-up mostly considered the HTHP and its respective parts.However, one must not forget that the flash tank was ordered and installed specifically for this experimental set-up.Hence, it was not included in the scaled-up dataset.Its mass was estimated to 1000 kg, including metal stabilization, pipes, electrical wiring and other auxiliary machinery.It is assumed to be of stainless steel, a low-alloyed material, modelled with the same ecoinvent process, as listed in Table 4.
For the final part of the model, water consumption must be considered.For the test operation, water was taken directly from the tap without further purification.As it is likely that a purification step is required for long-time operation a standard process for the provision of process water is used, and the amount reflects the steam mass flow, as Fig. 4. The scaling approach used with the BAMBOO HTHP added (Zeilerbauer et al., 2023).
shown in All flows accounting for the finalized LCI for 1 GJ of steam produced by the BAMBOO HTHP are presented in Table 6.Please note that, the total amount of material needed to build the HTHP, shown in Table 4, is divided by the total amount of steam energy produced (shown in the last line of Table 3) over the machine's life time in order to account for them evenly.
As a last step as the machinery has reached the end of its lifetime the deconstruction was modelled.Therefore, the different materials and their treatment were added to the model.For the sake of simplicity, we assumed that all of the metals were recovered and both types of polymers were treated in an incineration plant.For the working fluid, a reuse was assumed as many experts claimed that manufacturers are trying hard to preserve the working fluids for future applications.For the lubricating oil no treatment was modelled, due to the comparably small amounts and the assumption that it becomes used up over time.The waste treatment processes are displayed in Table 7.

Life cycle impact assessment (LCIA)
The results for the LCIA of the LCI presented in Table 6 are presented in Figs.6-11.The names 2 bar and 5 bar refer to the respective test points, whereas wind refers to generation of electricity by wind turbines and grid means the current electricity mix.LFO refers to light fuel oil.The geographical scope for all those inputs is Spain.For ADP elements, one can conclude that HTHPs do not provide benefits over any of the benchmarks, except for steam from biogas.The main reason is the demand for copper, which makes the largest contribution to HTHP manufacturing, with percentages of 60-80%, followed by electricity consumption.Individual contributions in absolute and relative values are provided in the Supporting Information.Because copper is practically indispensable in electronics, finding a lever is difficult.However, one explanation can be found for the cradle-to-gate system boundary.As End-of-Life was not modelled, no credits for recycling were awarded.In Western Europe, recycling rates of 48-65% are reached according to Henckens and Worrell (2020).The low values for the benchmarks can be explained by the relatively simple systems of boilers, which require less electronics and are mostly composed of steel, an abundant material.The comparatively high amount of ADP in the steam produced by biogas can be explained by farming, leading to organic waste input.Ramírez-Arpide et al. also found that conventional farming can lead to high ADP emissions (Ramírez-Arpide et al., 2018).
ADP fossil energy is much more straightforward, with very low values for renewable power and steam from biomass-fired boilers.Consequently, the Spanish grid yielded far higher values only to be surpassed by steam from natural gas and LFO.For the grid cases electricity demand for HTHP and circulating pump accounted for 97 and 98% of the total impact, respectively.As expected for heat pumps, the Spanish Grid converts around 350-600 MJ to 1 GJ of steam, while for 1 GJ of steam, considerably >1 GJ of fossil primary energy is needed.However, no analysis on exergy was performed for a more detailed review.In contrast to the first indicator, HTHP production has a negligible impact on the ADP-fossil energy impact category, highlighting again that for this category the energy needed is the driving force.
Acidification and Eutrophication are discussed together in Fig. 8 as they show the same trends.Biogas and Biomass exhibited much higher values for both categories than their fossil fuel counterparts.This is a common finding through LCAs by different authors and different scopes,

Table 6
Total LCI normalized to 1 GJ of Steam produced.Machinery data is presented over the whole 15 year lifespan, as the machinery needed is divided by the total amount of energy produced within 15 years.

Table 7
The waste treatment processes from the ecoinvent database used to model the End of life for the BAMBOO HTHP.featuring a benchmark between fossil and bio-based power/products (Chen et al., 2016;Fazeni-Fraisl and Lindorfer, 2022;Puschnigg et al., 2023).Like ADP elements, the high impact of steam from biogas can be found in the production of biogas, as organic waste comes mainly from industrial modern agriculture, which relies greatly on fertilization, herbicides, and other chemicals.Similar to the previously discussed results, the grid-case electricity is the most-defining step, with contributions over 95% again for the grid-powered system.For wind power, this figure is lower in percentage but considering the total impact, further optimization does not seem an intermediate target as also the best-performing benchmark, steam from natural gas, is outperformed by a factor of 5-10 roughly, depending on the desired steam pressure.In the same way as AP and EP the different toxicity-related categories were grouped together and are shown in Fig. 9.
For FAETP, the processes from dismantling the machinery make up the greatest part of the impact category, with the production of the machine being the second-most important part.All HTHP data is considerably higher than the benchmarks and in the extreme case between the 5 bar cases and steam from natural gas the results are >100 times as high.The detailed results reveal that the end-of-life treatment  for copper is responsible for over 96% of the total impact in all four HTHP cases.It has long been known that copper is toxic for marine organisms if concentrations are too high, as published by Gledhill et al. (1997) Neglecting the comparably extreme values for MAETP and TETP for steam from solid biomass, which were described in previous LCA studies (Alengebawy et al., 2022;Liang et al., 2012;Petrescu et al., 2016), it can be concluded that in this scope, HTHPs do not offer a benefit to conventional steam sources in terms of toxicity.
One important fact is that higher efficiencies yield significantly lower results.This becomes apparent when comparing the 2-bar to the 5-bar cases, with the 2-bar cases having much lower results because of the higher total heat output.Considerable efforts to lower these results are needed, considering that most process heat demand is provided via natural gas in the EU-28, as outlined by Rehfeldt et al., which showed the lowest results among the benchmarks (Rehfeldt et al., 2018).In general, it seems that for toxicity-related impacts, HTHP's performance already is worse than those of the most benchmarks; however, the most abundant steam source in Europe, natural gas, leads in terms of impacts.Thus, as no large levers were found within the individual process results, general optimization and improving efficiency seem to be the way to go forward.The Global Warming Potential (GWP) is presented in Fig. 10.
When comparing the two results for GWP in Fig. 10, it was found that they were very similar.Only small changes in the final numbers were found.In terms of climate change mitigation, the results clearly show that using the grid mix brings significant improvements when compared to the two fossil benchmarks, which are tremendously improved if renewable power is used.This finding is derived from the fact that in the grid cases, 95% of GWP can be attributed to the power used for the circulating pump and the HTHP itself.Thus, if a low-carbon mix is used, these results can be dramatically lowered, as presented above.In the favorable case of wind-powered HTHP (2-bar) in comparison with steam from LFO, 98% of CO 2 eq.can be avoided.One can deduce that HTHP's application helps combat climate change, and the results will only improve with an increased share of renewable power within the grid.Although these numbers are high, similar results have been reported in other works.Mateu-Royo et al. found a reduction potential of 98% in the GWP too for a Swedish case study when integrating an HTHP into district heating and comparing it to a natural gas boiler (Mateu-Royo et al., 2019a).
The results for the remaining two categories, Photochemical Ozone Creation Potential (POCP) and Ozone Depletion Potential (ODP), are presented in Fig. 11.For POCP, the HTHP tended to achieve better results than the benchmarks, except for the 5-bar grid case in comparison to natural gas.The main reason for the higher values obtained in the grid scenario is the amount of fossil fuels used, which leads to nitrous emissions that drive this value higher.The production of HTHP plays a minor role, with copper and steel being the main contributors.The case for ODP is much clearer; over 99% of the impact can be attributed to the nature of the WF.This is unfortunate because a proxy had to be used due to the lack of LCI data.However, the literature states that ODP for R-134a is approximately zero, and is still the defining step and magnitudes larger than the impacts of other parts of the HTHP, albeit at a very low level (Bolaji, 2010).As the same is said about R1336mzz(Z), one cannot answer this question in a conclusive manner (Molés et al., 2014).A  L. Zeilerbauer et al. likely reason is that one of the precursors in the production of R-134a contributes to this category and a part of it is leaked during production.

Sensitivity analysis -life cycle assessment
In the last step of this LCA, various parameters of this work were altered for sensitivity analysis.
First, the scale-up process of the HTHP system was evaluated.As previously discussed, a good fit was obtained using the linear regression shown in Fig. 5.However, as only one system was examined, no answers on possible errors could be made.Therefore, the HTHP's mass was altered by ±50%.For simplicity, the same procedure was performed for the flash tank.The two 2-bar cases are examined and the results are shown in Fig. 12.Both cases show a high sensitivity on changes in ADP elements and FAETP, followed by a medium sensitivity towards HTP and MAETP.This is because these categories are affected the most by the materials and power required for the production of machinery.Generally, the grid system is more robust to changes in machinery mass, as emissions are generally higher than those for the wind benchmark.One important finding is that when assessing only GWP, not much thought needs to be given to the scale of the mass, as its contribution is negligible in the grid-powered case.In the renewable energy scenario, the majority of midpoints are affected to a greater extent, as the overall impacts are lower, and the impacts of electricity play a minor role, thus increasing the importance of other components to the system.In the context of renewable energy, the impacts on most midpoints are more pronounced due to the overall lower environmental impacts, and the contribution of electricity to these impacts is relatively small.This highlights the significance of other components in the system and their impact on the overall environmental performance.
Another set of assumptions in this study pertained to the WF and its leakage.Although it was previously discussed that R134a was used as a proxy, new issues arose during the sensitivity analysis.The emissions to air were modelled using the ecoinvent flow "long time to air" for R134a.Upon reviewing the results, we found that the only midpoint categories affected by this flow were the two values for global warming potential.While this is unfortunate, the sensitivity for the GWP 100 was analyzed.We assumed 1% of annual leakage in our study, many other authors assumed a more conservative 5% (Jovet et al., 2022;Koroneos and Nanaki, 2017;Mateu-Royo et al., 2021;Sulaiman et al., 2022).Therefore, we tried 1, 5 and 10% to observe the results, shown in Table 8.Because of the much higher overall impact in the grid-powered case, the percentual changes are much smaller.One can deduce that 5 % of increased leakage corresponded to roughly 0.1 kg CO 2 eq.Keeping in mind that R134a was merely a proxy in this study, it should be considered that the new generation of working fluids features GWPs in single-digit numbers not close to 1000 kg CO 2 eq./kg, such as R134a or R245fa (1,1,1,3,3-Pentafluoropropane) (Sulaiman et al., 2022).
Therefore, these figures would very likely be much smaller in size than those presented here, and the results for the final results are also likely to be negligible.
Although other midpoint categories besides GWP could not be assessed via the leakage rate, it could be done indirectly.A higher leakage rate automatically leads to a higher need for refilling and, consequently, to increased production of WF.Thus, the amount of used and produced R134a (74.75 kg over the entire lifetime) was also altered by 75-100% and analyzed within the frameworks of the two 2-bar cases.The results are shown in Fig. 13.As expected from the previous results, WF production is not an issue for the majority of categories, except for ODP.The percentage changes show comprehensively that this indicator depends only on the working fluid produced.Considering the relatively low overall values for the 2-bar wind case, POCP is the only other category influenced by >3%, whereas the 2-bar grid case is basically not influenced at all, except for ODP.
Ultimately, the two most important parameters were shown previously.From the results in Fig. 6 to Fig. 11, it becomes obvious that the power supply for the HTHP and COP are the most influential parameters.Fig. 14 shows those results.From the 2-bar to the 5-bar case, COP was lower by a delta of 1.1.This corresponds to a rise of a little over 60% for the majority of midpoint categories.This view is very different from that provided by wind power.While the possible reductions often reach over 90% (AP, EP, GWP, and POCP), there are also somewhat smaller benefits, such as FAETP with 16% and HTP with 14%.ODP remains at the same level, and TETP even shows increased values by 56%.This highlights that improved efficiency, such as a higher COP as well as a generally different system, can lead to significant reductions in emissions.When comparing the two different COPs, it is important to keep in mind that these two levels may be equivalent in terms of thermal energy output; yet trivially, if a process requires 5-bar of steam then the more efficient 2-bar operation is not an option.These findings are consistent with the literature, confirming once more that renewable electricity offers enormous benefits to HPs and HTHPs in residential and industrial applications in general when it comes to their overall greenhouse gas emissions (Navarro-Esbrí et al., 2022;Riva et al., 2021).

Limitations, weaknesses and outlook
Finally, the limitations and weaknesses of this study are discussed.First, most of the data were obtained from the commercially available ecoinvent and Sphera databases.While it is understood that first-hand data allow for better and more precise results, the aim of this work was more conceptual and to guide other researchers or executives on how to assess an HTHP before investing in the technology.Similarly, much data depend upon estimates and the literature accessible to the public.
For the assessment of the BAMBOO HTHP system out of the 21 test points the ones with a higher COP were selected.The reason was that we aimed for a best-case scenario as HTHPs are already commercially available which reach higher pressures with higher COPs with a similar temperature lift.An example is the SPH ThermBooster which lifted water from 78 • C to saturated steam with a temperature of 159 • C (6 bar), and a COP of 3.0 (SPH Sustainable Process Heat, n.d.).
Upon assessing the inclusion of all relevant processes, one could argue that transport should have been modelled as well.The literature shows that, in LCAs dealing with energy-intensive processes, such parameters play a minor role and are often negligible.Moreover, no reallife implementation in the steel plant was realized; therefore, concrete transport modelling was omitted to allow for a more general model.
While a great deal of semi-empirical and pilot work focuses on WFs and their GWPs, this issue's urgency could not be reproduced in this LCA (Mateu-Royo et al., 2021;Molés et al., 2014;Navarro-Esbrí et al., 2022).Owing to the use of a proxy LCI-dataset and leakage only affecting GWP, an unprecise model was the result.Although the sensitivity analysis also did not show a significant influence on the case focusing on the status quo, this situation is unsatisfactory, and future research is still needed for more precise results.Although it should be emphasized again that R134a presents an extreme case with a very high GWP and the true emissions are very likely to be much lower.
Another issue with WFs is their fate after being leaked into the environment if not disposed properly.Hydrochlorofluorocarbons and hydrofluorocarbons, oftenly used as WFs in HPs and HTHPs are safe, but not stable in a chemical sense.They degrade quickly and the degradation pathways are well understood.A famous degradation product is trifluoroacetic acid (TFA).In fact, R134a degrades to TFA to 21% on a molar level and R-1234yf to 100% (Luecken et al., 2010).This strong acid and the corresponding salts are found all around the world, yet its origin is still disputed, whether is it occurring naturally or not (Joudan et al., 2021).As of early 2023 the European Chemical Agency (ECHA) has proposed to restrict per-and polyfluoroalkyl substances (PFAS), in the public discussion oftenly referred to as "forever chemicals" (European Chemical Agency, 2023).According to Consoli, WFs are included in the proposed restriction due to a number of them breaking down into TFA (Consoli, 2023).While there is no direct link to high toxicity for TFA, the substance is known to accumulate in water reservoirs and has shown impacts on the growth of microorganisms, thus leading to cautious legislative choices as presented before (Zhou et al., 2022).Therefore, precise analysis regarding the environmental impacts of working fluids and their degradation products is needed and those results need to be implemented to current LCA methods.

Determination of shadow price
The CML midpoint results presented in the preceding section were subjected to the damage factors listed in Table 1.The results are shown in Fig. 15.The Supporting Information lists the full table of the results including all data points.
When comparing the externalized shadow price, it becomes evident that 1 GJ of 2-bar steam produced by the wind-powered HTHP yields the lowest results of 0.5 €.This was expected, as this system showed the lowest contribution over an array of categories and exhibited an extremely low GWP.However, the shadow price also reflects problematic areas, such as the HTP and MAETP.The same is true for the windpowered 5-bar HTHP case, which incurs a higher cost owing to the lower COP.When comparing the grid-powered HTHPs, one realized that the more efficient 2-bar system has lower costs than steam from natural gas or biogas, which is also true for the 5-bar case, although the difference is almost negligible.This is interesting, as it shows again that efficiency optimization is a task that could be quite influential for the breakthrough of HTHPs.Natural gas and LFO for steam generation costs were mostly driven by their high GWP values, contributing 90% and 82%, respectively.Ultimately, the high cost of steam from solid biomass is mostly due to its extremely high MAETP value.Circling back to the introduction and the axiom that the government wishes to have the lowest shadow costs possible, as in the end, most of the cost will be externalized, and one can see that a subsidy for HTHPs could help mitigate shadow costs for the government and provide integration of HTHPs.Looking at the 2-bar wind case, a conservative subsidy of 3€ per GJ steam produced could be realized.Taking into account the 43.20 TJ produced over the lifetime of the plant, a subsidy of 129,600 € could be paid.Recent literature features 500-700 € of capital expenditures (CAPEX)/kW for HTHPs running over 8000 h annually, which would correspond to 125,000-175,000 € for the BAMBOO HTHP, when assuming the full capacity or 50,000-70,000 € for a realized steam output of 100 kW (Jovet et al., 2022;Saini et al., 2023).Either way, this subsidy could help reduce investment costs tremendously, rendering the technology more attractive for industrial companies.Obrist et al., found the same conclusion, as either high prices for carbon dioxide or incentives are needed to overcome investment barriers (Obrist et al., 2023).Saini et al. reported OPEX of 5 % of CAPEX for a lifetime of 15 years, while Jovet et al. used 2.5 % of CAPEX for a lifespan of 20 years (Jovet et al., 2022;Saini et al., 2023).This proves that CAPEX still incurs the largest cost over the entire lifetime.Ultimately, in the conclusion of their work Jovet et al., found that HTHPs are hardly profitable in the EU with current (May 2022) energy prices, again proving a need for government incentives (Jovet et al., 2022).

Conclusions
In this work, several findings on the topic of performing LCA for HTHP Systems were made.First and foremost, to our knowledge, we present the full LCA focusing on all areas of damage for HTHPs, not only GWP and/or ODP.From a life-cycle perspective, the GWP can be reduced with respect to any benchmark, depending on the power source.For example, the Spanish grid did not provide benefits in climate change mitigation when compared to steam from biogas, as opposed to wind power.Abiotic depletion (elements) was higher for all HTHP systems because of the increased material demand compared to simpler systems such as steam boilers.The acidification and eutrophication potentials were lower for HTHPs than for biomass as a steam source and at a level similar to that of fossil sources.The toxicity-related categories showed mixed results, with fluctuations between the HTHP operating modes and benchmarks, and no general remarks can be drawn.For most midpoint categories, the electricity used was the most important step for the HTHP cases, except for ADP fossil and FAETP.An exception was found in ODP and POCP, which showed a nearly 100% dependence on the working fluid's production and leakage.This was confirmed in the sensitivity analysis, which was performed because a proxy was used when modelling the refrigerant.Although much research has been conducted on developing new reagents with low GWP and how to use them, only a minor contribution to the GWP was found in this study.This was verified in the sensitivity analysis, in which the production and leakage were altered.It was discovered that there is a lack of LCI data for refrigerants in general and especially for state-of-the-art refrigerants.We hope that this data gap can be closed in the future, yet the authors think of this as a challenging endeavor as chemicals that can be used in HTHPs have a high number of possible production routes.Moreover, not only the production but also degradation pathways and the resulting pathways due to formation of polyfluorinated substances such as trifluoroacetic need to be investigated in future works.Our scaling approach was also subjected to a sensitivity analysis simulating an overestimation as well as an underestimation of 50% of the HTHP mass.ADP elements and FAETP were affected the most, whereas other categories were less likely to be affected.In the case of renewable power, the results were lower overall, and higher percentual changes were recorded.Generally, the decarbonization of the electricity supply was found to have the greatest influence on the results when the Spanish grid and wind power were compared.Calculating the shadow price for different HTHP systems revealed possible amounts of subsidization under the premise that external costs will have to be paid by the public to mitigate the negative effects these emissions will have in the future.

Fig. 1 .
Fig. 1.The high temperature heat pump system with flash tank during the test operation.

Fig. 2
Fig.2.Schematic overview of the high temperature heat pump system.In contrast to a regular water-to-water high-temperature heat pump this setup employs an extra circuit composed of circulation pump, flash valve and flash tank to generate steam.A portion of the pressurized hot water flow, driven by the circulation pump, is expanded and evaporated at the flash valve.Steam can then be drawn from the flash tank while the unevaporated part of the flow is recirculated.

Fig. 3 .
Fig. 3.The operating points evaluated in course of the test runs.The three selected points represent different saturated steam output pressures.Graphic adapted from Riedl et al. (2022).

•
ES: Process steam from biogas (Process Data Set: Process Steam from Biogas 90%; Technology Mix Regarding Firing and Flue Gas Cleaning; Production Mix, at Heat Plant; MJ, 90% Efficiency (En), n.d.) • ES: Process steam from solid biomass (Process Data Set: Process Steam from Biomass (Solid) 90 %; Technology Mix Regarding Firing and Flue Gas Cleaning; Production Mix, at Heat Plant; MJ, 90 % Efficiency (En), n.d.) • ES: Process steam from natural gas (Process Data Set: Process Steam from Natural Gas 90 %; Technology Mix Regarding Firing and Flue Gas Cleaning; Production Mix, at Heat Plant; MJ, 90 % Efficiency (En), n.d.-a) • ES: Process steam from light fuel oil (LFO) (Process Data Set: Process Steam from Light Fuel Oil (LFO) 90 %; Technology Mix Regarding Firing and Flue Gas Cleaning; Production Mix, at Heat Plant; MJ, 90 % Efficiency (En), n.d.)

Fig. 5 .
Fig. 5.The linear fits of all IEA data (in blue) and the lower values between a heat supply capacity of 30 and 700 kW (in red).COD: Coefficient of determination.

Fig. 8 .
Fig. 8. Acidification and Eutrophication Potential categories for 1 GJ of steam produced by the BAMBOO HTHP.LFO: Light fuel oil.

Fig. 9 .
Fig. 9.All toxicity-related midpoint categories for 1 GJ of steam produced by the BAMBOO HTHP.LFO: Light fuel oil.

Fig. 10 .
Fig. 10.Global Warming Potential with and without biogenic carbon for 1 GJ of steam produced by the BAMBOO HTHP.LFO: Light fuel oil.

Fig. 11 .
Fig. 11.Photochemical Ozone Depletion Potential and Ozone Depletion Potential for 1 GJ of steam produced by the BAMBOO HTHP.LFO: Light fuel oil.

Fig. 14 .
Fig. 14.A chart depicting the changes throughout the midpoint categories from the 2bar grid base case.Lines do not indicate a correlation; they are shown as visual aid.ADP: Abiotic depletion potential, AP: Acidification potential, COP: Coefficient of performance, EP: Eutrophication potential, FAETP: Freshwater aquatic ecotoxicity Potential, GWP: Global warming potential, HTP: Human toxicity potential, MAETP: Marine aquatic ecotoxicity potential, ODP: Ozone depletion potential, POCP: Photochemical ozone creation potential, TETP: Terrestrial ecotoxicity potential.

Table 2
Measured and calculated values for the system during the test operation.COP: Coefficient of performance.

Table 3
Power needed for steam production and upgrading and characterization temperatures for the steam obtained.COP: Coefficient of performance.

Table 4
Life Cycle Inventory for the HTHP system scaled linearly from Ecoinvent 10 kW System.

Table 8
Sensitivity analysis pertaining to the global warming potential (GWP) for the two 2-bar cases with different percentages of WF leakage.