Comparative LCA studies of simulated HMF biorefineries from maize and miscanthus as an example of first‐ and second‐generation biomass as a tool for process development

5‐Hydroxymethylfurfural (HMF) is a versatile platform chemical for a fossil free, bio‐based chemical industry. HMF can be produced by using fructose as a feedstock. Using edible, first‐generation biomass to produce chemicals has been questioned in terms of potential competition with food supply. Second‐generation biomass like miscanthus could be an alternative. However, there is a lack of information if second‐generation lignocellulosic biomass is a more sustainable feedstock to produce HMF. Therefore, a life cycle assessment was performed in this study to determine the environmental impacts of HMF production from miscanthus and to compare it with HMF from high‐fructose corn syrup (HFCS). HFCS from either Hungary or Baden‐Württemberg (Germany) was considered. Compared to the HFCS biorefineries the miscanthus concept is producing less emissions in all impact categories studied, except land occupation. Overall, the production and usage of second‐generation biomass could be especially beneficial in areas where the use of N fertilizers is restricted. Besides, conclusions for the further development of the on‐farm biorefinery concept were elaborated. For this purpose, process simulations from a previous study were used. Results of the previous study in terms of TEA and the current LCA study in terms of environmental sustainability indicate that the lignin depolymerization unit in the miscanthus biorefinery has to be improved. The scenario without lignin depolymerization performs better in all impact categories. The authors recommend to not further convert the lignin to products like phenol and other aromatic compounds. The results of the contribution analyses show that the major impact in the HMF production is caused by the auxiliary materials in the separation units and the required heat. Further technical development should focus on efficient heat as well as solvent use and solvent recovery. At this point further optimizations will lead to reduced emissions and costs at the same time.


| INTRODUCTION
The platform chemical 5-hydroxymethylfurfural (HMF) has been attracting the attention for several years. From 2016 to 2021, more than 1100 articles were listed in Google Scholar where "HMF" and "bioeconomy" were mentioned together. HMF is known as a sustainable and versatile platform chemical for the future fossil-free chemical industry and lists as one of the most promising bulk chemicals of a biobased future (Bozell & Petersen, 2010). Bicker et al. (2003) already stated that HMF is a "sleeping giant" of chemical intermediates. Thirty years later, HMF is still in its infancy. There are only a handful of companies offering HMF on the market. MarketWatch (2023) lists eight companies as major players. Most of them are still in the development and pilot stages, so prices for sulfurfree HMF are high and not very competitive. Especially in the area of bulk chemicals and high-value chemicals, the bioeconomic transformation is faltering. According to "Insights into the European market for biobased chemicals", only 0.3% of platform chemicals produced in the EU-28 are biobased (Spekreijse et al., 2019). Many technologies on the way to commercialization do not get beyond pilot operation in technology readiness level (TRL) 4-6, the so-called valley of death (Kampers et al., 2021). Afterwards, there is often a lack of investment to further scale and drive the development. Venture capital in the chemical sector has been stagnating in Europe for years. In Germany, it is even declining (Rammer et al., 2022). McKinsey sees a growing trend that the topic of sustainability has arrived in the chemical and manufacturing industries and among their investors. This will also lead to increasing profits and market shares in the long term. A strategic benefit of the topic of sustainability is therefore given (Ezekoye et al., 2022). But there is a need of scalable sustainable production processes (Dutta et al., 2012). Teles (2019) also addresses this issue and concludes that there is a lack of a holistic view of sustainability along the entire process chain and thus, despite a large number of publications, proof of better sustainability is lacking. He focuses on the fact that most publications end with analyzing data from laboratory tests and thus the concept of sustainability is limited to the use of renewable raw materials and the selection of catalysts and solvents. Therefore, this work will give information about the Hohenheimer miscanthus biorefinery concept to produce HMF and comparing it with one of its competitors. This is not other fossil-based bulk chemicals, but fructose-based HMF produced via the same conversion routes.
There are some publications dealing with the environmental impact of HMF-based products but these studies are barely presenting values for HMF itself (cf. Davidson et al., 2021). Eerhart et al. (2012) for example reported −0.09-0.33 kg CO 2 -eq kg −1 product basket, meaning an undefined mixture of HMF, HMF esters, levulinic acid and its esters. However, information about the sustainability of the production processes and HMF itself varies. The numbers differ widely when the Global Warming Potential (GWP) according to ISO 140040 is used as the calculation benchmark. Ranges between 326 and 1.159 kg CO 2 -eq kg −1 HMF are reported (Schöppe et al., 2020). The reasons for this are manifold and show how the difficulty to compare quantitative results from different studies.
Some possible reasons for these wide variations are different system boundaries, allocation methods, and study frameworks, but also different technical processes for HMF synthesis and purification. Davidson et al. (2021) give a good overview here. For example, in Eerhart et al. (2012) high-fructose corn syrup (HFCS) is converted to HMF using methanol and diluted sulfuric acid, and another study evaluated a process in an H 2 O/ DMSO system and using a 20 wt.% Sn on γ-Al 2 O 3 solid catalyst (Kougioumtzis et al., 2018;Schöppe et al., 2020). Other scientists work with ionic liquids, for example 1-ethyl-3-methylimidazolium chloride, [EMIm]Cl, or acetone-water mixtures (Graham & Raines, 2019;van Putten et al., 2013). In addition, the quality and reproducibility of the raw data may also vary. Therefore, Schöppe et al. (2020) criticized that many of the processes for HMF production presented in literature do not provide robust, detailed data. Götz et al. (2022) recently published a modular, smallscale plant concept of hydrothermal HMF production in aqueous acidic reaction media, including a technoeconomic analysis (TEA). With this data made available for the conversion of the lignocellulosic biomass miscanthus, the environmental impact of the process can now be analyzed. HMF is obtained from fructose via dehydration, whereas lignocellulose contains cellulose and thus polyglucans. This leads to a lower HMF yield compared to fructose as feedstock (Steinbach et al., 2018;Świątek et al., 2022;Tan-Soetedjo et al., 2017).
Besides the bulk chemical HMF, another platform chemical, furfural, can be produced from hemicellulose and lignin is also a product of high interest for the bioeconomy. The lignin can be further depolymerized, also via hydrothermal processes to aromatic monomers (e.g., phenol) and oligomers (Götz et al., 2022;Schuler et al., 2019). Besides this, biorefineries for material usage also produce renewable energies. In the presented biorefinery concept the aqueous residues can be used in a biogas plant, to produce the needed heat for the hydrothermal processes, or in a wastewater treatment plant like simulated for this life cycle assessment (LCA). With fructose as a feedstock only HMF and a minor amount of hydrochar is produced during the presented processes (single purpose biorefinery).
The use of first-generation feedstock as input materials, for example, fructose from maize starch has been questioned for their environmental sustainability and potential competition with the global food supply (Humpenöder et al., 2013). In contrast to nonedible lignocellulosic secondgeneration biomass, first-generation biomass is defined as mostly edible parts of agricultural crops, like carbohydrate stored from corn or sucrose (Haji Esmaeili et al., 2020). Second-generation feedstock such as the perennial crop miscanthus is considered promising alternatives.
Unlike first-generation energy crops like maize, the soil is covered for the entire cultivation period after the miscanthus establishment, which reduces the risk of soil erosion and supports soil carbon sequestration (Harris et al., 2015). Typically soil carbon increases ranging between 0.7 and 2.2 t carbon ha −1 year −1 are reported for miscanthus cultivation on arable land (McCalmont et al., 2017). Miscanthus effectively relocates nutrients from the above-ground biomass to the below-ground plant parts after the vegetation period (Cadoux et al., 2012). Furthermore, it has been reported that miscanthus can provide an additional habitat for a number of species, contributing to an increased biodiversity potential of agricultural landscapes (Lask et al., 2020;Werling et al., 2014). Socio-economically beneficial characteristics include reduced management requirements (relative to annual crops) and stable biomass yields during the cultivation period as well as versatile utilization options (Lewandowski, 2016;McCalmont et al., 2017). In addition, miscanthus can be cultivated on marginal or contaminated lands, on which biophysical conditions prevent an economically viable cultivation of conventional food crops (von Cossel et al., 2019).
For these reasons, miscanthus is considered a promising alternative to first-generation feedstocks. This leads to the question if the production of the platform chemical HMF from the second-generation biomass miscanthus in the evaluated multi-output biorefinery ( Figure 2) performs better in terms of environmental sustainability from a life cycle perspective than the production from firstgeneration biomass in a single output biorefinery following the same methods and processes.
In this study first-generation feedstock will be fructose originated from maize-based HFCS. This is the benchmark biomass for the HMF on the market today (Kläusli, 2014). In Menegazzo et al. (2018) it is stated that the current commercial production of HMF is originated from syrups made out of energy crops. A patent research made by Rosenfeld et al. (2020) found, that "most of the examples given in the patent literature were done with fructose as feedstock".

| Goal and scope
The goal of this LCA is to compare the environmental performance of biobased HMF production from two sources. The benchmark scenario based on HFCS biorefinery located in Southwest Germany is subdivided into two further cases (Scenario A and B, Table 1). This enables a sensitivity analysis of the syrup production site. Once the HFCS production is integrated into the T A B L E 1 Description of the different biorefinery scenarios studied in this work. biorefinery (Scenario A) and once the syrup is produced in Hungary and then transported to the biorefinery (Scenario B). In 2020, Hungary was the largest exporter of HFCS in Europe, with Germany being one of the largest importers in the world (Observatory of Economic Complexity, n.d.). This suggests that a HFCS biorefinery would source feedstock from Hungary. The basic scenario of the alternative, the Hohenheim biorefinery concept using the example of miscanthus (Scenario C), first splits the biomass, as described in more detail later, to then produce the main product HMF and the by-products furfural, charcoal, and a mixture of phenolic aromatics in several process sub-steps. In addition, hotspots, the materials and therefore the process sub-steps with the highest impacts, in the HMF production from miscanthus are identified. As an example, the LCA of a possible optimization approach is then evaluated in a second iteration, which fits identically to the optimization proposals of the previously conducted techno-economic assessment (Götz et al., 2022). This study and the results of the LCA showed that the economic viability of the miscanthus biorefinery benefits from not implementing the phenol separation module (M4 in Figure 2). The modular concept ( Figure 2) of the biorefinery makes this possible without having to make major changes or adjustments to the process (Section 4.4). For this reason, this setup was also tested in this study (Scenario D, Table 1). Scenario E is a sensitivity analysis of the allocation methods comparing economic allocation with mass allocation methods.

Abbreviation
The system boundaries were set to cradle-to-gate, addressing the raw materials, production, and manufacturing stage to investigate possible optimization potentials for further process development. The investigated product system includes the cultivation of miscanthus and grain maize as well as their conversion to HMF (Figure 1). The study does not include the emissions of the commissioning and decommissioning of the plant itself.
The miscanthus cultivation took place in the southwest of Germany at the research station "Unterer Lindenhof" of the University of Hohenheim near the city of Reutlingen. The biorefinery concept is a small-scale and regional onfarm biorefinery approach to close regional nutrient cycles and offer new business opportunities for small-scale farmers in rural areas. These smaller farms are often found in the selected region. It includes soil preparation, crop establishment, fertilization, weed control, and harvest of the biomass. For miscanthus, a 20-year cultivation period is taken . Biomass yields were assumed in accordance with the geographical focus and account for natural yield variability.
Maize cultivation was simulated in Southwest Germany in the same region as the miscanthus field trials for better comparability and in Szolnok, Hungary. Hungary is one of the top exporters of HFCS. In terms of the general temporal scope the study is based on the current state of knowledge with regard to biomass cultivation systems and biomass conversion systems for a consideration of future commercial biorefineries.
F I G U R E 1 Simplified product system for HMF production from miscanthus and maize using the example of the Baden-Württemberg scenarios. 1 Lignin for miscanthus-based HMF production without the process sub-step for phenol separation (cf. Götz et al., 2022). Both scenarios: one including the depolymerization of lignin to phenolic mixtures and one without, leading to only lignin as product is evaluated here. HMF = 5-Hydroxymethylfurfural; HFCS = High fructose corn syrup. A detailed listing of all processes, including waste treatment, can be found in the supplementary materials.
Information for the miscanthus conversion to HMF was mainly derived from the AspenPlus®-simulation described in Götz et al. (2022), which details a small-scale miscanthus biorefinery with a biomass intake of 500 kg fresh matter (FM) h −1 and an output of 52 kg HMF h −1 (7′000 operating hours year −1 ) as the main product. Two different concepts of the miscanthus biorefinery are tested as described later in Table 1 and named Scenario C and D.
In both concepts the by-products furfural (62 kg h −1 ) and char (340 kg h −1 ) are generated. But one biorefinery approach ends with lignin as a product and in the other approach a further conversion of the lignin into aromatic compounds is linked. Thus, they include different process steps and differ in the type and quantity of auxiliary materials. Part of the system evaluation at the end is thus a discussion of the process engineering concept of the multi-output biorefinery. More details about the differences and process sub-steps are given in Götz et al. (2022). Similarly, data for the HMF production from HFCS were derived from an AspenPlus®simulation. The inventory of HFCS production from grain maize was modeled according to Broekema and Kramer (2014). In order to avoid an imbalance in the comparison between the HFCS biorefinery and the miscanthus biorefinery due to different standards in energy and auxiliary material supply, it was assumed that heat and, if possible, auxiliary materials in all processes are renewable (e.g., ethanol from fermentation or heat and electricity cogeneration from wood chips).
The functional unit of this LCA is defined as 1 kg HMF available at a refinery in Southwest Germany. The authors chose economic allocation by default (see Section 4.1) and calculating allocation by mass in Scenario E to express the differences due to different methods. The calculation methods are given in the supplementary materials. The used prices are presented in Table 2. In total, five different configurations of a biorefinery producing HMF are compared (Table 1).
All assessments were conducted for six midpoint impact categories from the impact assessment method collection ReCiPe 2016 v1.1 (Huijbregts et al., 2017). This includes GWP (100 years), terrestrial acidification (TA), particulate matter formation (PMF), and land occupation (LO). These categories were selected based on preliminary results of this study, where they have been identified as contributing to at least 80% of the total impacts of the default miscanthus-and maize-based product systems on endpoint level (ecosystem quality and human health). In addition, freshwater eutrophication (FE) and marine eutrophication (ME) were considered due to the relevance of eutrophication concerns associated with agricultural processes (Bechmann & Stålnacke, 2016;Kirschke et al., 2019;Wagner & Lewandowski, 2017).
Calculations of the environmental impacts were conducted using the LCA software Brightway2 (Mutel, 2017). In order to account for natural variability (triangular distribution) a range of biomass yields and carbon sequestrations (miscanthus only) were assessed for the miscanthus and maize cultivation systems (Tables 3 and 4). This was done by conducting Monte Carlo analyses with 1000 calculations providing ranges for the results in the assessed impact categories.

| Primary biomass production
Miscanthus cultivation over a period of 20 years requires a range of agricultural procedures, including soil preparation, establishment, fertilization, harvesting, and recultivation of the field. Inventory data for these agronomic procedures including the method used for direct land use change, were taken from Lask, Rukavina, et al. (2021). Field emissions, including that is nitrous oxide, nitrate and phosphorus as well as heavy metals were modeled as suggested in Zampori and Pant (2019). Yield ranges shown typical for miscanthus cultivation in Southwest Germany were taken from own field trials (Weik et al., 2022). These are complemented with estimates on typical soil carbon sequestration associated with the cultivation of miscanthus on arable land. Carbon storage for the duration of the cultivation period only was considered as suggested by Lask, Rukavina, et al. (2021). On the one hand, the parameter carbon sequestration can easily dominate the variation in GHG emissions associated with miscanthus production. On the other hand, the values for carbon sequestration can range broadly from 0 to 2.2 t C ha −1 and year −1 . To avoid that this parameter impedes the further assessment, a conservative approach of 0 t C ha −1 and year −1 was considered. In this study, the application of no nitrogen fertilizer to miscanthus is considered. This is reported for commercial practice (Jason Kam, personal communication, October 2020 and McCalmont et al., 2017). Key parameters of the inventory of miscanthus cultivation are summarized in Table 3.
Inventory data for the cultivation of maize were taken from the Agri-footprint® LCI-database (version 5.0). The inventory for maize grain production in Germany was adjusted. This reflects maize yield variability in the German region of Baden-Württemberg (BW) using information from Statistisches Landesamt Baden Württemberg (2022). The same approach was taken for maize production in Hungary using information from FAO stat (, 2022). Table 4 summarizes the key inventory parameters of the maize cultivation, representing the yield variability between 2005 and 2020.

| Biomass conversion of miscanthus
Figure 2 describes the biorefinery system boundaries of the actual manufacturing stage of the miscanthus biorefinery (Scenario C). In a hydrothermal process miscanthus will be split in its lignin, cellulose, and hemicellulose fractions. The hemicellulose and the amorphous cellulose will be hydrolyzed to sugars. The hemicellulose reacts further to furfural (module 2) and the cellulose-based glucose will be converted to HMF (module 3). Both reactions are hydrothermal reactions. The residues, the waste water, will be disposed. There are two possible ways to do so. It can be disposed in a biogas plant (Götz et al., 2022) to close the nutrient cycle via applying the resulting digestate on the fields and for an internal supply of heat. An alternative is to dispose the harmless process water containing of slightly acidic water with some side products (c.f. Wüst et al., 2020) and unreacted sugars in a wastewater treatment plant. This option has been chosen in favor of its flexibility. It shows that the concept also works without a biogas plant available. Another aspect is, that the development of an effective coupling of the biorefinery and the biogas plant is currently ongoing in the research project BIOKOP and robust data are not yet available. Therefore, also the nutrient recycling is excluded from this study. In Scenario D the lignin from module 1 will not be depolymerized. This is a result of the high emissions of module 4 and shows one possible way to integrate LCAs in process optimization loops. The procedure to convert miscanthus to platform chemicals in a hydrothermal green chemistry approach is described in detail in (Götz et al., 2022). The raw data on the production process were received from small pilot plant trials at TRL 5-6 with a scale of 2 kg FM miscanthus per hour in the biorefinery pilot plant at the Agricultural Research Station Lindenhöfe of the University of Hohenheim and have partly been published (cf. Świątek et al., 2022). At this stage of development, all unit operations have already been identified and transferred to apparatuses. However, the individual modules of the process chain are usually still run separately and not in sequence in order to simplify the execution of experiments. In production plants, this is the case to implement cascade effects, material flow recycling, and optimized heat management. Consequently, this leads to less use of materials and higher conversion rates. To illustrate this, this interlocking process chain is simulated using the simulation software AspenPlus®.
With the help of the common simulation package AspenPlus®, conversion calculations can be coded and linked into the simulation package as custom function blocks, enabling proper calculations of the mass, and energy balances. As a main property method NRTL was used to provide a simulation of the complete manufacturing stage. The complexity of the blocks can easily be adapted to the information available, and the compound-based description can be varied between blocks. Cascade effects and heat exchanger management, which can strongly influence the environmental impacts, were simulated with the software. As a result, the TRL along the entire process chain has to be aligned and increased to TRL 7. For this purpose, a number of assumptions had to be made, as described in Götz et al. (2022). In order to provide a realistic F I G U R E 2 System boundaries of the multi-product biorefinery with the processing modules (M1-M5). M1 to M5 = modules/process sub-steps of the biorefinery. Aux.1&2, Activated Carbon and Ethanol; Aux.3, Ethyl acetate; Misc., miscanthus; KOH, potassium hydroxide. insight into the current state of the art, these assumptions are kept conservative and leave some room for further process development on the way to TRL 9. All the made assumptions in this previous study are valid for the current one. An excerpt of it is provided in the supplementary materials.
For the life cycle inventory (LCI), the inputs and outputs are shown in Table 5. The values are normalized to 1 ton of miscanthus FM h −1 . A commercial scale would be between 4.5 and 5.5 tons per hour, or about 35,000 tons of miscanthus FM h −1 , according to calculations by Götz et al. (2022). The study does not include the emissions of the construction and decommissioning of the plant itself. Therefore, the emissions caused by different scales does not affect the results of this study.

| Biomass conversion of maize
Starch production from maize grain was modeled using information from the ecoinvent 3.7.1 database. In line with the scope of this study, the default dataset 'maize starch production' was adjusted by changing heat supply from conventional to renewable source and taking country-specific electricity supply (German or Hungarian grid mix, respectively). The following production of HFCS (42% fructose-55% glucose) from maize starch was modeled using inventory information from Broekema and Kramer (2014). Similar to the inventory of the maize starch production, the inventory was adjusted by selecting heat and electricity supply in line with the study's scope. For HFCS production in Hungary, transport distance of 1000 km was added to account for the transport of the HFCS to Southwest Germany (default: Szolnok ➔ Stuttgart, min = 700 km, max = 1200 km). Key inventory of the HFCS production process is presented in Table 6. Detailed inventory for all conversion processes is given in the Supplementary Material.
The further simulation of the intermediate product HFCS to the desired product HMF in the biorefinery concept evaluated here is also carried out via AspenPlus®. Methodologically, the approach is identical to the miscanthus biorefinery. However, there are some changes. Because a processed, sugar-containing feedstock is used, several modules of Figure 2 are omitted. For example, biomass preparation (M1) is no longer necessary. Furfural synthesis (M 2A) and purification (M 2B), as well as the processing of the lignin (M 4) are also eliminated, since the feedstock contains neither hemicellulose nor lignin. Thus, less heat and a lower temperature level of 200°C instead of 210°C is required. The amount of auxiliary material used for purifying the HMF will increase, because more HMF is produced per m 3 feedstock, and less water, reaction medium, is removed from the system through T A B L E 5 Major in-and outputs of the HMF production from miscanthus including the lignin depolymerization (Scenario C). The values for the miscanthus biorefinery without the further depolymerization is given in the Supplementary Materials.

Item
Quantity Unit Note: The values have been rounded, which results in a mass balance that is not exactly balanced.
*The wastewater is the amount of water that need to be purged per hour and it contains also unreacted sugars and side products.
by-products. At the end, HMF and biochar are produced. Furfural, lignin, or a mixture of aromatic compounds are not formed. In Table 7 the LCI data for the HMF production from maize-based HFCS are summarized.

| Comparative assessment (economic allocation method)
Figure 3 (and Table S14) describe the resulting emissions calculated by economic allocation (Scenarios A to D). The median for GWP of Scenario C is 2.23 kg CO 2 -eq kg −1 HMF. Slightly higher, 2.64 kg CO 2 -eq kg −1 HMF, are the emission of the HFCS-BW biorefinery (Scenario A). If the cultivation and processing of the maize to HFCS is located in Hungary, the footprint increases. Without the module for splitting the lignin to aromatic compounds (Scenario D) the emitted CO 2 equivalents can be reduced by half compared to Scenario C. For GWP box size and whisker length are the same in all scenarios. For LO, the results of Scenario C are similar to those of the HFCS Scenario B. As with the GWP, the environmental impact is reduced compared to Scenario C if the lignin is not further depolymerized. However, the lowest LO was achieved in this study for the HFCS Scenario A. The mean values are between 3.7 and 5 m 2 × a kg −1 HMF across all scenarios. The interquartile range is different compared to the GWP. The highest is observed in the Scenario B and the lowest interquartile range is seen in the Scenario A.
In the PM impact category, emissions are also lowest after the miscanthus biorefinery simulation without the further refinement of lignin. This is similar to the GWP. In this subgraph, the occurrence and magnitude of outliers in the HFCS scenarios is notable. Note: The values have been rounded, which results in a mass balance that is not exactly balanced.
Abbreviation: HFCS, high-fructose corn syrup. a The initial sugar solution had 2 wt.% fructose. But these calculations are based on a feedstock containing 55 wt.% glucose and 42 wt.% fructose. b The initial amounts of water and acid to reach a dilution of 2 wt.% is higher but in this table only the amount of water and acid that need to be replaced is considered. c The wastewater is the amount of water that need to be purged per hour and it contains also unreacted sugars and side products.
The FE potential of the Scenario C is similar to the potential for Scenario B. The HFCS Scenario A produced the biggest impact. The option without the lignin splitting reduces PO 4 3− equivalents by about another 0.1 g per kg HMF compared to Scenario C. The clearest differences between the use of HFCS and miscanthus as biomass in a biorefinery with the primary objective of producing HMF can be observed in the categories TA and ME. Here, the expected emissions from both miscanthus options are estimated to be far lower than in the HFCS scenarios. Table 8 shows the quantitative values for the medians underlying the Scenario A across the assessed impact F I G U R E 3 Life Cycle Impact Assessment (LCIA) results for the comparison of miscanthus-and maize-based production of 5-hydroxymethylfurfural (HMF). Scenario A: HFCS-based HMF using maize produced in the state of Baden-Württemberg; Scenario B: HFCS-based HMF using maize produced in Hungary. Scenario C: miscanthus-based HMF including the depolymerization of lignin; Scenario D: miscanthus-based HMF without the further depolymerization of the lignin to aromatic monomers. categories and the results for a second allocation method, the physical allocation, are given. Due to the change of the calculation method, GWP calculated by physical allocation dropped to approximately 20% of the result of the monetary allocation shown in Table 8. It is the same for the other categories PM, TA, and LO. The ratio between the values of the two scenarios is not always the same. The values for Scenario C are median values including uncertainties due to the variance in harvestable yield and CO 2 -sequestration. Scenario E was calculated only for one single case. Those uncertainties have different impacts on impact categories.

| Contribution analysis
The question of how to further reduce emissions in the context of technology development to achieve industrial maturity of lignocellulosic biorefineries for HMF production is addressed with hotspot analyses of the Scenarios C and D. These are being compared with the actual standard biomass for HMF production, HFCS. Since the miscanthus for the simulations of this study is grown in BW, the reference is Scenario A. Figure 4 shows the proportional environmental impacts of the biorefineries including CO 2 sequestration. A classification of emissions into the categories biomass cultivation, transportation, HMF production raw materials, HMF production energy, and waste management can be found in the supplementary materials. In all scenarios transportation is neglectable because of its regional concepts in favor of a circular bioeconomy approach.
In the GWP category, emissions from ethyl acetate (consumption and disposal) have the largest share of 35%. Ethyl acetate is used in the purification process sub-step of the aromatic compounds (Scenario C). As an organic solvent ethyl acetate is disposed separately. Subsequently, the consumption and disposal of ethanol, biomass supply, and heat contribute 23%, 12%, and 11% to the GWP emissions of Scenario C. In the impact categories PM and TA, the heat and the ethanol used for HMF purification  are the largest polluters. The impact of ethanol is only 13% if we take a closer look at the impact category LO. Here, miscanthus cultivation with 48% and heat with 38% account for the majority of the remaining emissions. In terms of eutrophication potential, the impact of biomass cultivation on freshwater resources is the highest, and the impact of the ethanol has the highest share of the ME potential.
Even without lignin splitting (Scenario D), the percentual share of the impact in the categories TA, LO, FE and ME remains relatively the same. The largest impacts originate from the previously described input materials. Most of the emissions in Scenario D are distributed among the ethanol input, the process heat and the miscanthus biomass, since in the Scenario D ethyl acetate, potassium hydroxide and the disposal of ethyl acetate are not present. The share of ethanol for the process sub-step HMF purification thus increases for the GWP from 23% in the Scenario C to 39% in Scenario D.
Through the comparison of Scenario D and A, it is noticeable that the share of the initial biomass production along all impact categories is higher for Scenario A. In addition, ethanol consumption is again one of the main factors. In contrast to the miscanthus scenarios, the provision of heat plays a subordinate role, but a higher influence of the use of activated carbon (AC), used for the purification of HMF , can be seen in the categories GWP, PMF, and TA.

| Discussion of the used allocation methods
The biorefinery produces several products, so emissions needed to be apportioned between HMF and charcoal, or also furfural and lignin/phenolic compounds. This allocation could be done on a mass basis. However, for future commercialization an economic partitioning based on market prices would be preferable (Eerhart et al., 2012). The authors chose economic allocation by default because economic relationships reflect the socio-economic demands that lead to the existence of these processes and their products, or multi-function systems, in the first place (Ardente & Cellura, 2012;Azapagica & Cliftb, 1999). The authors also concur with Peereboom et al. (1998) arguing T A B L E 8 Impacts for physically allocated (Scenario E) and monetary-allocated (Scenario C) HMF produced from miscanthus including lignin depolymerization. that economic allocation better reflects the societal cause of emissions (Ardente & Cellura, 2012). HMF is the main product of interest, but the yields are relatively low. Byproducts like the resulting hydrocarbons, take a larger share by mass. Physical allocation would place more emissions on (undesirable) by-products. However, the need of this biorefinery is to produce HMF, not hydrochar. Guinée et al. (2004) discuss that "prices can fluctuate considerably, but shares often remain fairly constant, especially over the longer term" (Ardente & Cellura, 2012). By the fact that the previous TEA calculated minimum sales prices, it is hardly possible to operate at lower HMF prices. The price is still very high compared to fossil base chemicals, so that in the next years higher prices could be badly enforced. The furfural price of 0.8 EUR kg −1 in Table 1 is at the lower end of the price range between 2006 and 2016 and will probably continue to rise in the future (Krishna et al., 2018). Since furfural and HMF are chemicals for partly the same applications, it can be assumed that the prices paid for HMF will increase to a similar extent. The ratio would therefore hardly change. According to the ISO standards, allocation should be avoided by dividing the unit process into subprocesses or by system extension (International Standard Organization, 2006). This cannot be implemented because further subdivision of the process is not possible. The simulation already represents the unit operations, the smallest units in process engineering, in the manufacturing stage, the focus of this work. System expansion is not plausible, especially for the Scenario D, since this is not currently produced on a large scale. In addition, the possible applications are so diverse that different values could be credited depending on which chemical or material is replaced by lignin. For example, lignin can replace lithium in batteries or phenol in resins (Baloch & Labidi, 2021;Rodriguez Correa & Kruse, 2018). The choice of allocation method has a strong influence on the quantitative results of a HMF biorefinery. The economic allocation of the Scenario C results in 20% higher values in all categories than the physical allocation of the same scenario (Section 3.2). The share of HMF in the annual turnover is used as a basis on the one hand and the mass of HMF in relation to the total mass of all products on the other hand (see supplementary materials). By changing only the allocation method within the same study and raw data, huge differences result. Since the feedstock miscanthus is a hardly soluble solid, the char and lignin formation is preferred compared to HMF formation. Those solid particles can be converted to HTC char by a solid-solid reaction and less HMF is formed (Kruse et al., 2016). This decreases the emissions to be imputed to F I G U R E 4 Contribution analysis for both HMF-miscanthus biorefineries producing HMF, Furfural, char and aromatic compounds (Scenario C) and lignin instead of aromatic compounds (Scenario D), compared with HMF-HFCS biorefinery with maize cultivation in Baden-Württemberg. Included is biomass supply, transport and all major auxiliaries to produce HMF. FE, Freshwater eutrophication; GWP, global warming potential; HMF, 5-Hydroxymethylfurfural; LO, Land occupation; ME, Marine eutrophication; PMF, Particle matter formation; TA, Terrestrial acidification.
the HMF under Scenario E. However, the main purpose of this biorefinery is HMF production. In this context, the LCA results are always a snapshot. As a result, the authors are not focusing on discussing the quantitative LCA results with literature but on the question of how the benchmark Scenario A can be further optimized. For example, by using miscanthus as feedstock?

| Differences between maize (HFCS) and miscanthus as a feedstock
Based on the results for GWP, PMF, and LO (Figure 3), the question of whether HMF production from firstgeneration biomass, using the example of maize-based HFCS, causes higher environmental impacts than from second-generation biomass, using miscanthus, especially Scenario C, cannot be answered unequivocally. The lower impact of biomass cultivation does not have such a strong effect on the LCA results for the final HMF product in the above-mentioned impact categories. Even though the process of HMF production from miscanthus is more complex than from maize-based HFCS, the differences in the conversion process are not the main driver for the unexpected results. Growing 1 ha of miscanthus is more environmentally friendly than growing 1 ha of maize, but this is not as clearly reflected in the environmental impact of the product.
The mechanism of producing HMF has been largely understood. HMF is formed from the dehydration of fructose (Jung et al., 2021;Körner et al., 2018). 9.8 tons ha −1 is an average yield for maize in BW between 2005 and 2020 (Destatis, 2022). A maize grain consists on average of 60.6 wt.% FM total sugars in the form of starch and high amounts of sucrose, glucose, and fructose (Zabed et al., 2016). Thus, carbohydrate yields of up to 5.9 tons ha −1 can be achieved. After the processing of maize into HFCS, 42 wt.% is in the form of fructose. In comparison, the biomass yields of miscanthus are at 15.6 tons ha −1 . Due to a cellulose content of 44.3 wt.% FM (Götz et al., 2022), the carbohydrate yields of up to 6.9 tons ha −1 . For miscanthus this is only available in the form of cellulose. However, glucose forms a natural equilibrium in aqueous solution with fructose, which can then further react to HMF. The isomerization of glucose to fructose represents the rate-determining step (Steinbach et al., 2018). This results in lower HMF yields from cellulose and other glucose-based biomasses in the literature compared to fructose-based biomasses. Świątek et al. (2022) report HMF yields from miscanthus biomass of 19.8 wt.%. Her experiments with fructose show HMF yields of up to 39.6 wt.%. However, isomerization can be accelerated using process technology by, for example, the use of catalysts (Delidovich & Palkovits, 2016;Steinbach et al., 2018;Yan et al., 2017). Thus, more HMF per hectare can be produced from maize than from miscanthus. The emissions are spread over more products. However, the advantage of lower greenhouse gas emissions of 0.05 kg CO 2 -eq kg −1 miscanthus compared to 0.31 kg CO 2 -eq kg −1 maize does not fully affect the end product assessed here to the same extent.
Since, maize yields in Hungary are about one third lower than in BW, GWP emissions are the highest. For the same amount of HMF, more cultivation area and thus indirectly more primary resources are needed in Hungary. The influence of the low yield in the cultivation area of Hungary can be seen directly in the category LO. Thus, the most important factors for the relatively small difference in some impact categories between maize and miscanthus (Scenario C) are the higher conversion rates and yields from maize-based HFCS compared to miscanthus biomass, using an inefficient process sub-step (see Section 4.4).

| Trade-offs and the role of fertilization
A general recommendation for or against first-generation cultivated biomass in HMF production cannot be made. Especially LO performs better using maize. But other impact categories perform better in the miscanthus biorefinery concepts. The reason for the lower performance of the Scenario A in the category TA lies in the data set used for maize cultivation. In BW fertilization is carried out with a larger amount of pig manure. This causes ammonia outgassing, contributing to an increase in acidification potential. Also, in the category FE, the fertilization strategy causes the impacts to be highest for the Scenario A. The lower performance of this scenario can be explained by an increased use of phosphorus in the used data set, compared to Scenario B. Overall, fertilizer use in the Germany is higher than in Hungary and higher overall for maize cultivation than for miscanthus cultivation. This leads to higher leaching, outgassing, and emissions in general, but also to higher yields.
Those fertilization strategies for maize will in future be limited on many areas due to regulations. It leads to yield losses and therefore losses in sales, but also lowers the emissions. The losses could only be minimized by undersowing or adapted crops and crop rotations, but probably not compensated completely. Growing miscanthus is another alternative to farm productively on these areas. The results show that there is an additional offer that HMF biorefineries can provide in addition to producing high-value chemicals and products. It also offers a solution to a major challenge facing farmers in future. This is not primarily a technology development issue, but a business development issue and will definitely help to overcome the valley of death mentioned in the introduction. These findings also contribute to the further development of the process and its transformation from research to business.
Fertilizers are not applied effectively enough and be utilized insufficiently by the plants, which leads to high environmental pollution. Half of all nutrients applied worldwide cannot be taken up by plants, thus polluting soil, air, and water (Lassaletta et al., 2014). Steffen et al. (2015) already assumed that global nitrogen and phosphate emissions have already exceeded the stress limits of ecosystems. To fight this, the European Commission has prepared the EU's Nitrates Directive. This is an instrument of European water protection with the aim of keeping nitrate pollution of aquifers and reservoirs in the EU below a limit of 50 mg nitrate L −1 . Despite some advantages, member states including Germany continue to exceed the limit values. In the reporting period from 2016 to 2018, about 27% of the monitoring sites (mainly agricultural areas) still exceeded the limit value (BMEL and BMU, 2020). Other countries/regions like the Netherlands, Denmark, northern France, and Spain have similar problems (Wiegmann et al., 2020). In Germany, the Fertilizer Regulation therefore continues to be made stricter. The subsequent amendment to the Fertilizer Regulation provides for an overall reduction of the fertilizer requirement in particularly polluted areas by 20%. Among other things, this is also a challenge for crop rotations emphasizing maize, as required for the maize-based HMF production. Actually, high yields can then only be achieved with soils that have a high subsequent supply of nutrients and have been treated with organic fertilizers for many years. On average, quality losses and lower yields are to be expected. A comparison of the HFCS biorefinery in BW and Hungary provides an outlook on the consequences of reduced fertilizer use in maize cultivation for the emissions of an HMF biorefinery. The falling yields per hectare cause GWP, PMF, and LO to increase. This explains the trade-off between less land occupation and high nitrogenand phosphorus-related impact values in categories like TA, ME, and FE.

| Value engineering and evaluation of the biorefinery process
In the scenarios presented, it has been shown that, depending on the prioritization of the impact categories, Scenario C is advantageous compared to Scenario A and B, even though HMF yields are higher for Scenario A. Nevertheless, the question arises whether it is possible to develop the process in a way that the negative environmental impact of the production of chemicals is reduced further. For the further development of the biorefinery concept, similar conclusion can be drawn from this work as from the TEA published by Götz et al. (2022). The congruence of the results of the TEA and LCA is a major advantage for the road to commercialization of the technology. A Pareto optimum is not yet recognizable. There is currently no conflict of interest between improved sustainability and reduced manufacturing costs.
The optimization steps should be focused on advancing the biorefinery concept presented in the Scenario D without further inefficient process sub-steps to convert the lignin into aromatic compounds. The lignin depolymerization consumes a lot of energy and chemicals. Also, this process has a lower TRL compared to the other process steps (Götz et al., 2022). The authors recommend to demonstrate the biorefinery at first without this depolymerization module (Scenario D). Further scale-up activities in the field of lignin splitting will help to lower the emissions of Scenario C and make it an attractive concept. The use of auxiliary materials has to be reduced, and the product yields and recovery rates of the material and energy streams have to be improved. Another option is to end with the product lignin (Scenario D). On the one hand, lignin is a transportable product, so that it could be processed centrally (outsourcing from the company balance sheet), and on the other hand, high-quality products are currently being developed directly from lignin (Section 4.1). In an iterative approach the outcome has been validated in terms of sustainability (Scenario D). Besides the above discussed differences, all impact categories have in common that the impacts of the Scenario D are lower than in Scenario C. One of the main reasons are the differences in TRL.
The heat management systems and efficient heat exchangers are a key issue in thermochemical processes and the use of ethanol in the separation processes must be minimized or the recycling rate has to be increased. Finding another suitable green solvent is hardly possible. Acetone is a very good solvent for HMF (Saliger, 2013). However, the biorefinery would have to be explosion proof, and the GWP is 2.8 times higher at 2.33 kg CO 2 -eq kg −1 acetone. Since renewable energies and biobased ethanol as a solvent in the HMF-separation unit have already been used, sustainability can only be improved through efficiency improvements and more efficient recycling loops. The amount of ethanol to be used in the separation unit to leach the HMF from the activated char should be decreased further. This could be done by an optimized design and operation of the adsorption column, or the recovery process should be improved to lower the losses. Here the simulation made relatively conservative assumptions according to actual laboratory and small pilot plant data and 5 wt.% of the ethanol is purged to avoid the accumulation of impurities. Currently, we use a vacuum distillation in the pilot plant to purify the HMF. In a first iteration a DWSTU block was used for the AspenPlus®-simulation leading to a high bottom temperature, leading to higher energy consumption but also in HMF losses due to slight degradation and polymerization reactions. A thin-film evaporator (perhaps a multi-stage approach is needed) would allow the temperature to be lowered to some extent. This reduces the heat requirement, increases the purity of the HMF, and allows higher recycling rates of the ethanol. To simulate this, a FLASH3 block was used to separate HMF from the solvent. It was able to reduce the energy demand. But no further reduction of the purge stream (5 wt.%) or higher HMF purities were assumed for the assessment. Future work has to be done to validate the lower energy demand and test, if with these improvements the quality of the HMF can be increased and if changes in the apparatus for purification can lead to savings in auxiliaries and energy. The use of auxilliaries and energy are the most important impacts on environmental sustainability.
The costs of heat provision could also be reduced by using cheaper energy sources, like internally sourced energy. Recycling the unreacted carbohydrates and the side products in a biogas plant to produce energy is not the only benefit of coupling the biorefinery with a biogas plant. The heat from the CHP unit, that is not (optimally) utilized might be also an example for cheaper heat sources. A cascade utilization via the heat exchangers of the biorefinery could be implemented. To optimize the heat recovery process in the biorefinery the research project BIOKOP funded by the German agency FNR will investigate and model the heat flows between a biogas plant and an HMF biorefinery. To improve costs and sustainability at the same time reducing consumption and improving the heat recovery should be preferred. Also, the soluble inorganic compounds in the miscanthus biomass will end in the digestate to fertilize the soils in future. A sustainable biorefinery should include also those aspects.
After the HMF separation the pretreatment module is the second largest energy consumer. Currently, a singlestage reactor system was designed for this purpose. Here, water, miscanthus, and acid are heated to 210°C under evaluated pressure. This happens at the beginning of the process chain. The largest mass flows in the system are brought to the highest temperature. This consumes a lot of energy. To optimize this, a multi-stage reactor system for miscanthus decomposition is to be developed and tested. Under hydrothermal conditions, the hemicellulose is hydrolyzed to xylose at lower temperatures (180°C) (Bobleter, 1994). The xylose is withdrawn and the lower mass flow is heated to 210°C in a second reactor to hydrolyze the cellulose. The stepwise increase in temperature and reducing the heat demand at the highest temperature level through reduction of the mass flow, energy saving will be obtained. In a follow-up project, the process simulation from this work is currently being adapted to validate the new concept.
Furthermore, future process development will lead to a longer utilization period of the adsorbent used in the purification of HMF, the AC. In the carbon materials research focus, the Department of Conversion Technologies of Biobased Resources is also working to develop biobased, sustainable ACs for the process. This can further reduce emissions. In the case of the GWP in Scenario D, the use of AC is the fifth largest emitter accounting for 14% and a factor to be considered. Currently fossil AC was used in the model. It was the only dataset available in the ecoinvent databank. The influence of the used hard coal AC on the GWP is 7.79 kg CO 2 -eq kg −1 AC. Using for example coconut shells as a feedstock could reduce its emissions to 1.1 kg CO 2 -eq kg −1 AC without any possible reactivation (Vilén et al., 2022).
Producing more HMF decreases emissions and costs. As mentioned in Section 4.2 glucose from miscanthus can be isomerized to fructose. However, the cost efficiency of heterogenous catalysts and energy-and chemical-demanding process conditions are remaining challenges (Yu et al., 2019). Each isomerization step also causes additional emissions, so the authors do not consider this option.

| CONCLUSION AND OUTLOOK
By providing insights in the HMF-biorefinery concept and studying the environmental impacts this study is another small puzzle piece to give more information about the environmental sustainability of HMF. The question of whether the production of HMF from second-generation biomass, using the lignocellulosic biomass miscanthus as an example, causes less environmental impact than HMF from HFCS, a first-generation biomass, can be answered on the basis of the various scenarios. At the current stage of research and development, Scenario D should be preferred to all other tested scenarios. The first iteration of the LCA led to the suggestion not to convert the lignin at all. For further optimizations the authors suggest to priorities: 1. The optimization of the HMF-separation unit. A thinfilm evaporator and methods to reduce the needed amount of ethanol should be considered. 2. A multi-stage pretreatment module to reduce the amount of heat needed for splitting the miscanthus. 3. Developing biobased ACs suitable for long usage in HMF adsorption.
Do these conclusions apply to other first-and secondgeneration biomass? Generalization is not possible, but there are some indications, especially for HMF biorefineries, that it is also true for a wider range of stalk lignocellulosic biomasses and processed carbohydrates from first-generation biomass. The lignocellulosic biorefinery results in more valuable products. Other types of lignocellulosic materials are mostly by-products of food or feed production. This reduces the emissions that the biorefinery must account for. Carbohydrates can be converted to HMF without purification, and as we have seen, the separation unit consumes most of the energy. Secondgeneration biomass also requires less fertilization.
In future, for reasons of water and nature conservation, there will be agricultural areas where traditional crop rotations with maize can no longer be cultivated economically. Here, potentials for miscanthus cultivation arise. The relocation of the nutrients to the below-ground plant parts, allows the miscanthus to remobilize the nutrients in the following vegetation period, contributing to an efficient nutrient use. In combination with an extensive root system this also reduces nitrate leaching from decaying biomass. With the cultivation of miscanthus and its processing into high-quality platform chemicals, farmers also gain new opportunities and alternatives for profitable cultivation on land where the use of fertilizers has been restricted due to environmental protection measures and should be reduced. By developing circular and modular biorefinery concepts, as described in this study, new business models to convert second-generation biomass from areas, where other high yielding crops would require a lot more fertilizers, into platform chemicals for a numerous of applications are being created. It follows that farmers will take on additional tasks in the future. In addition to providing food, feed, and bioenergy, it is now also possible to produce valuable and sustainable commodities for the chemical industry in regional cycles. These key aspects of the bioeconomy and agro-industrial biorefineries can contribute to a lasting and sustainable development of agriculture and rural areas. Second-generation biomass HMF can play an important role to make the transformation from an economy driven by fossil resources to a bioeconomy without fossil bulk chemicals happen.