Allocation in recycling of composites - the case of life cycle assessment of products from carbon fiber composites

Our study includes the cut-off approach and the end-of-life recycling approach, because they are both frequently used in LCA (Frischknecht 2010; Nordelöf et al. 2019). Frischknecht (2010) claims that these approaches complement each other: the cut-off approach reflects a strong interpretation of sustainability because uncertain benefits of future recycling are not included in the calculations, whereas the end-of-life approach reflects a weaker interpretation of sustainability as it accounts for the expected benefits of future recycling.

In addition to these two approaches, the Circular Footprint Formula (CFF) (European Commission 2018) was selected because it is part of the Product Environmental Footprint (PEF) framework, which we expect to be used by many LCA practitioners in the future. The PEF framework has been introduced by the European Commission in order to (among other things) increase the harmonization of environmental assessments of different products, services, and organizations (Manfredi et al. 2012).

A brief overview of the three selected approaches is found in Table 1; they are described in more detail in Sects. 2.1.1–2.1.3. Variables used in the three approaches are found in Table 2 and those common for the three approaches are illustrated in Fig. 1. Different guidelines suggest that different allocation approaches are to be used in different situations; these suggestions and how our case study applies the three allocation approaches are discussed in Sect. 4.3.

As composites in fact consist of several material components that have been merged into a new material, a recycling process will often involve a separation (or deconstruction) process with potentially multiple useful materials as outputs. This means that the recycling process introduces an allocation problem of its own: allocation between the different material components (i.e., “co-products” of the recycling process). The same is true for any deconstruction process in the life cycle of complex products (e.g., the dismantling of a car) as well as for any raw material processing that generates multiple products (e.g., in oil refineries or biorefineries). If the recycling process can have a significant contribution to the total impacts, it is important how this allocation is done. For composites, successfully recycling the material components typically requires considerable efforts, e.g., a large energy input, and how allocation is done may therefore have a significant impact on the LCA results.

Dealing with the allocation problem for the separation process may require that the components of the composite are modeled as individual materials (e.g., a polymer matrix and carbon fibers in the case of carbon fiber composites) rather than a single composite material, as functional and quality changes may differ between the components. When this is needed, the considered recycling allocation approaches must be adapted to allow for consideration of multiple materials with different fate. Equations for adapted recycling allocation approaches were therefore developed and these are described in Sect. 2.3. The impacts of the recycling process are allocated to each material i using the allocation factor α_i, which can be based on, for example, the mass or energy content, or economic value of the materials. For more discussion on approaches for co-product allocation, see Hermansson et al. (2020).

Equations 4–6 show the adapted allocation approaches, where each composite component is handled separately. Note that the equations only describe the impact that is allocated to a considered composite material that is recycled and that can be part of a product life cycle. Other materials and various other activities, like the product use, might have to be added in an LCA. The variables that have been added or changed compared to what is presented in Table 2 are provided and explained in Table 3. Note that the composite recycling is now conceptualized as a separation or deconstruction process.

Equation 4 shows the adapted cut-off approach.

Equation 5 shows the adapted end-of-life recycling approach.

Equation 6 shows the adapted version of the CFF.

The case study is used here for illustration of the influence of applying different allocation approaches. Therefore, a fictitious product made from only primary composite material was chosen and one such product weighs 1 kg. A passive product type is considered to allow for the simplification that the use phase can be neglected. Several other simplifications were also made as described below. The functional unit was the service provided by one piece of fictitious product φ used for σ years.

It was assumed that the product was manufactured, used, and discarded in Germany. Any transports and collection of waste were excluded from the study as these were deemed to be negligible in relation to the impacts of the other processes.

For simplicity, the use-phase is also assumed to have negligible impact (i.e., E_U = 0) and, hence, can be excluded from this particular case study. This adequately represents the use of the composite in a passive product, such as a man-hole cover, but not in an active product such as a vehicle.

All data were, if not stated otherwise, taken from the Ecoinvent APOS 3.3 database (Wernet et al. 2016). All LCA modeling was done using the software OpenLCA v.1.10.

The polyacrylonitrile (PAN) precursor fiber production data was based on a dataset by Fazio and Pennington (2005) in the European reference life cycle database (ELCD), which was modified to fit the Ecoinvent nomenclature. Carbon fiber production inventory data were taken from Romaniw (2013). Data for PA production was approximated by a nylon 6–6 production process.

Three different recycling methods were considered: grinding (Case A), pyrolysis (Case B), and supercritical water treatment (Case C). Data for the recycling processes were taken from Dong et al. (2018) who base their data on the work by Hedlund-Åström (2005), Witik et al. (2013), and Knight (2013). It was assumed that when recycling is not done, the discarded product is sent to landfill. The modeled systems are shown in Fig. 2. Details on the modification of datasets and on system modeling can be found in the supplementary material.

Grinding of a composite can in some cases separate the polymer matrix and the fiber (Dong et al. 2018). In this study, however, for illustration purposes, the composite is ground in Case A and the resulting powder of mixed material is assumed to replace only polymer in a secondary composite material application.

In pyrolysis (Case B), high temperatures are used to decompose the polymer matrix and to liberate the carbon fibers. The energy carrier is assumed to be electricity (Dong et al. 2018). Clean fibers are recovered, and thermal energy and/or secondary fuels or chemicals can be produced (Yang et al. 2012). For example, Cunliffe et al. (2003) write that the polymer matrix can be turned into oil and gas products. In this study, the decomposition of the polymer is assumed to generate a chemical comparable to petroleum, and therefore substituting the production of petroleum. Note that we assume that 100% of the polymer is decomposed into an oil; in reality, this value is likely lower.

The process used in Case C is supercritical water treatment. The energy carriers are assumed to be electricity and heat generated from natural gas (Dong et al. 2018). The inventory data used are for the recycling of a composite with epoxy matrix, but in our case study, the composite contains a thermoplastic polymer; this was not corrected for, but is not expected to have influenced the results in important ways. It is assumed that the recovered fibers can substitute fibers with a quality very close to primary fibers, and that the recovered polymer substitutes a similar polymer produced from fossil resources with negligible quality degradation and losses in the recycling process.

The impacts of the recycling process were allocated between the composite material components based on mass for comparability and data availability reasons. As a simplification, it is assumed that none of the recovered materials in the three cases needs any (intensive) processing before entering the second product life cycle. This is more likely to be true for the fibers, but the polymer matrix would need some processing before the next application as the polymer has likely been degraded or contaminated to a certain extent depending on the recycling method used. However, as this need will also vary depending on the secondary application of recycled materials, which in our case study is not specified, it was excluded from further assessment.

This study considered the environmental impact categories of energy use and climate impact. The selection was based primarily on data availability, but it was also deemed sufficient for the purpose of the study as it serves to illustrate how the choice of allocation approach influences the results. In addition to this, as climate impact often correlates well with other emission-based impact categories, it can be seen as a proxy for other impacts (Janssen et al. 2016). The impact assessment methods used were the cumulative energy demand and GWP100 in CML2001 as provided by Ecoinvent 3.3 (Wernet et al. 2016).

The changes in the modeling of the composite into several parallel material flows were proven useful in the modeling of recycling of a composite. The conceptual split of the composite into different material flows may seem obvious once it is established, but a composite is not necessarily thought of as several material components that will have different fates in the end-of-life process, as well as in subsequent product life cycles. If the composite had been modeled as a single material at recycling, the differing properties, quality degradations, and supply and demand of the different components would not have been captured. We argue that this approach for assessments that include recycling of composites is useful and likely applicable also to many other types of composites, allocation methods, and recycling cases, than the ones used here, but we have not tested this.

The adapted allocation approaches, described in Eqs. 4–6, were applied to three different cases of recycling of a product manufactured from CFRP. The resulting climate impacts and energy uses are seen in Figs. 3 and 4. Note that the results are based on information found in literature and databases and are generated for the purpose of illustration of the influence of the allocation approaches on the cradle-to-grave environmental impact only. The modeling was not done with any specific real product in mind.

The cut-off results are identical for all recycling options. This means that the cut-off approach gives no incentive to improve recycling processes. In fact, it hardly gives an incentive to send the product to recycling at all. The only difference between no recycling and the cut-off approach is that the disposal part in the cut-off approach is slightly smaller as parts of the product’s materials are recycled into a second life cycle. The impacts related to the disposal process are small, and this difference is hardly visible in Figs. 3 and 4. The cut-off approach does, however, give incentives to use recycled materials in product manufacturing (see Eqs. 1 and 4). Adding recycled input as a parameter in this case study would have made the modeling of different scenarios complicated and difficult to interpret and we deemed this to be beyond the scope of this study. However, a simplified calculation shows that recycling components as in Case C and using them in a secondary product application would result in both the climate impact and energy use of the secondary product being reduced with around 50% compared to using only primary materials (i.e., half of what is presented in Figs. 3 and 4). This approach thus strongly rewards using recycled materials in composite-based product manufacturing. The use of recycled materials in product manufacturing should be considered in future research on recycling allocation; this is particularly important in prospective studies and for composites that already today include recycled materials.

The end-of-life recycling approach results in lower climate impact and energy use for Case C compared to Cases A and B. This is because of the relatively large credits associated with the avoided production of fibers and polymers due to the high quality of the recovered components. In Case A, the carbon fibers do not substitute other carbon fibers but instead a polymer and the product is therefore awarded the relatively smaller credit corresponding to replacing polymer. In Case B, the polymer is degraded and assumed to replace petroleum production, which gives a smaller credit than when the recycled polymer is assumed to replace polymers. Note that the larger reduction awarded due to substitution in Case C is partly counteracted by the relatively high impacts of the supercritical water recycling process. Otherwise, the climate and energy advantage of Case C would have been even greater. This illustrates how some allocation methods reward recovering high-quality materials from the product, and thus highlights the importance of capturing this in the allocation procedure. It also shows that the end-of-life recycling approach gives not only an incentive to recycle the product, but also to prioritize the recovery of that material for which the primary material production has the highest environmental impact. What the method does not do is to give any incentive to use recycled materials in production of composites (see Eqs. 2 and 5). The strong influence of what is substituted and of the considered quality degradations in this method highlights the importance of relevant input data. Accurate data are particularly difficult to obtain when modeling future recycling systems, in this case, the pyrolysis and supercritical water treatment processes. In such cases, key parameters could be varied in a sensitivity analysis.

When using the CFF, the climate impact and energy use results strongly depend on the supply and demand balance of the market (i.e., allocation factor A). For example, in Case A (grinding), the fibers are not reused; instead, the powder containing both polymer and fiber is reused and replaces only polymer matrix in a new composite. This means that the output of recycled polymer material is higher than the input, as it is diluted with milled fibers. As a consequence, the climate impact and energy use results are lower for Case A compared to Cases B and C when there is a low demand for recycled carbon fibers. This result is intuitive as it suggests that when the demand for recycled fibers is low, it is better to replace the polymer than the fiber. With a high demand for recycled fibers, though, the CFF results indicate that fibers should be recycled separately, particularly if the recycling process does not degrade the fibers significantly (for Case C). Hence, the CFF gives incentives not only to provide high-quality recycled materials to the next product life cycle (just like the end-of-life recycling approach) but also to recycle and recover materials for which there is a demand on the market. This is true even if the recycling process itself has a high environmental impact. Note that the degree of complexity of the CFF makes it particularly challenging to apply in studies of future recycling systems, as it considers both recycling rate, the market supply and demand balance, what is substituted and quality changes of each component in recycling. As society changes and technology develops, all these variables are expected to change. A way to manage this in future-oriented studies could be to look into best- and worst-case scenarios.

What is clear from the results in this study is that the different allocation approaches provide different incentives for recycling. Differences in incentives provided by different recycling allocation approaches have been discussed before in the literature on a more generic level, for example, by Frischknecht (2010) who relates the cut-off and end-of-life recycling approaches to different views on sustainability (see Sect. 2.1). Frischknecht (2010) writes that it is not possible to claim that one approach is more correct than another, but that both approaches can be rational depending on one’s value system.

It is not possible to provide general guidance on which allocation method is more relevant as this depends on the context and purpose of each LCA study. There are, however, some recommendations in the literature: BSI (2011) stipulates in PAS 2050:2011 that the cut-off approach is to be used if the recycled material does not keep the same inherent properties as the primary material, and the end-of-life recycling approach is to be used if the recycled material does keep the same inherent properties as the primary material. Likewise, the Greenhouse Gas Protocol by WBCSD & WRI (2011) stipulates that the end-of-life recycling approach is to be used in processes where the recycled material can be used to replace primary materials with the same inherent properties, and the cut-off approach is to be used in open-loop recycling systems which have recycled material inputs and outputs. In some cases, both the end-of-life recycling approach and the cut-off approach could be relevant. In such a situation, the Greenhouse Gas Protocol (WBCSD & WRI 2011) suggests that the cut-off approach is more appropriate to use when the product contains recycled input but the product is not recycled, in cases where the markets for recycled materials are saturated, when the use phase of the product is long or uncertain, and when the content of recycled materials in the product is influenced by the companies’ activities alone. The end-of-life recycling approach, on the other hand, is to be used when the recycled material content of the product is unknown, when the market for the recycled material is not saturated, and when the use phase of the product is short or well known.

When attempting to apply the guidelines by the BSI (2011) and WBCSD & WRI (2011), described above, to our three recycling cases, it can be argued that the changes in quality of both fibers and polymer mean that the cut-off approach is more suitable for Cases A and B, whereas the end-of-life recycling approach is more suitable for Case C. However, it could be argued that today’s market is not saturated for recycled carbon fibers, and the end-of-life recycling approach could be seen as the most appropriate approach for Case B as well. This highlights again the importance of the choice of allocation approach in future-oriented studies, and ideally, both the cut-off and the end-of-life recycling approaches could be used to illustrate the potential effect of changes in both market saturation and technology development, where, for example, quality degradation could decrease with continued technology development over time.

Section 4.2 highlights the importance of the allocation factor A in the CFF, which allows for accounting for relevant market mechanisms in the modeling of recycling. This inclusion allows for a more accurate modeling of recycling in cases when the market for the secondary material is sufficiently established for its bottlenecks to be identified. When there is not yet an established market, the PEFCR Guidance stipulates A = 0.5 (European Commission 2018). This makes the CFF a compromise between the cut-off and the end-of-life approaches. As an alternative, different values for A can be applied to illustrate the uncertainty, as in our case study, where different assumptions on the future demand for recycled carbon fibers were included (see Sect. 3.3.2). Note, though, that the PEFCR Guidance allows for 0.2 < A < 0.8 only, which does not cover the full range of possible situations. This full range is better illustrated by the two extreme approaches: cut-off and end-of-life recycling (see Figs. 3 and 4).