LFP-based binder-free electrodes produced via fused filament fabrication

Carbon coated-LiFePO4 (LFP) is a strong candidate as lithium-ion battery (LiB) cathode due to the combination of safe operation, stable electrochemical performance and positive environmental impact as does not depend on Co, which is toxic and a critical raw material. In this work, we report the development of binder-free LFP cathodes fabricated by fused filament fabrication (FFF) technology. Several novel carbon-LFP filaments have been developed to 3D-print LiB cathodes, analysing both the carbon to LFP ratio in the formulation and also the impact of the carbon source used as current collector, i.e. glassy carbon (GC) microspheres or carbon black (CB), in the electrochemical performance. LFP remained stable upon debinding and sintering at temperatures as low as 500 °C as determined by x-ray diffraction. The conductivity of 3D printed LFP monoliths was 2.06 × 10−4 S∙cm−1 at 50 °C, which is fairly close to that of LFP produced via conventional processing. This is mainly attributed to the preservation of the carbon coating around the LFP particles after debinding and sintering under controlled Ar atmospheres. The LFP-based electrodes containing CB or GC microspheres as conductive additives exhibited specific capacities of 150 mAh g−1, and over 95% coulombic efficiency after 100 cycles, showing no significant performance losses. These results largely exceed the performances reported for previous LFP-based electrodes produced via FFF as the non-active binder is removed upon fabrication.


Introduction
Nowadays, rechargeable lithium-ion batteries (LiBs) are the most widely used energy storage devices due to their outstanding energy and power density, only comparable to fuel cells. Apart from their capacity to provide energy, LiBs are highly appreciated because of their reliability, long-term cycle life, negligible memory effect and low cost. Therefore, they are the state-of-the-art in applications such as portable electronics, power tools and medical instruments. Furthermore, they have been considered as the best choice to boost pure and hybrid electrical vehicles and are at the forefront of storage technologies for renewable energy resources [1,2]. Nevertheless, LiBs must face several challenges to consolidate their position in those applications, as is the ever-increasing demand of higher energy density, extended cycle life, safety and cost-efficiency [3].
To address these concerns, two different strategies have been typically explored. On one hand, a great deal of effort focuses on improving both the electrochemical performance and stability of active materials and the search for novel compositions, as is the case of solid electrolytes, which are in the spotlight as they promise a radical advance in terms of safety [4,5]. On the other hand, performance enhancement may be achieved via manufacturing routes to maximise the volumetric and gravimetric energy density of the final devices [6,7].
Conventional LiBs electrodes fabrication is mainly based on the preparation of a ceramic slurry, which contains the active material, polymeric binders (e.g. PVDF), organic solvents and conductive agents. Then, thin layers are spread onto current collectors via coating techniques such as doctor blade or slot-die. This manufacturing approach results in robust and reliable batteries, although there exist some limitations, mainly related to the use of non-electrochemically active additives up to 50 vol%, decreasing dramatically the volumetric energy density [8], and also to the reduced number of battery geometries, i.e. cylindrical, coin, prismatic and flat cells [9]. In the last decade many alternative strategies have emerged in the search for surface area enhancement in LiBs electrodes, including chemical vapour deposition, wet and dry etching or sputtering, among others [10][11][12][13][14][15][16][17].
Fused filament fabrication (FFF) 3D printing employs solid thermoplastic-based filaments that do not demand special preservation requirements to ensure printability. This combined with the ease of use, capability to print large parts, possibility of multimaterial printing [39] and the low cost of both 3D printing systems and materials, justify the research efforts towards the development of LiBs using this technology [40]. Nevertheless, to date, the number of 3D-printed electrochemical applications via FFF is rather limited. Graphene/polylactic acid (PLA) composite filaments were recently marketed and tested to produce 3D printed LiB anodes [41,42]. Analogously, Reyes et al developed LTO and LiMn 2 O 4 -based filaments with only 4-6 vol% of active material along with carbon additives (16-24 vol%) and PLA as binder (70-80 vol%) to produce 3D printed electrodes for wearable electronics [43]. Similarly, LFP cathodes and LTO/graphene anodes [44] and LFP/PLA electrodes [45,46] have also been reported. However, the electrochemical performance in the previous cases can be considered as modest as they exhibit capacities/energy densities far from the theoretical values, mostly due to the large amount of inactive polymer in the filaments, which in turn could lead to electrode delamination and performance losses as a consequence of secondary reactions [47][48][49]. Moreover, those filaments cannot be used to produce full ceramic electrodes, which would hamper their use in high energy density applications [50][51][52].
Regarding FFF 3D printing of full ceramic LiBs electrodes, LiCoO 2 and LTO have been recently produced based on a patented procedure [53]. The 3D printed full ceramic electrodes showed high and competitive electrochemical performance, i.e. 129 and 168 mAh·g −1 respectively, and efficiencies of ≈95%, which is fairly close to the state-of-the-art electrodes obtained by conventional uniaxially pressing [54]. As LFP has been considered as a strong candidate in the development of batteries for applications such as electric vehicles due to their stability during lithiation/delithiation, low cost and safety [55][56][57], in the present work we report the fabrication of novel binder free LFP/C cathodes via FFF 3D printing, including studies of the phase stability and characterisation of the micro-and macrostructure, electrical conductivity and electrochemical performance tests of the optimised electrodes.

Filament fabrication
Filaments of carbon-coated LFP and LFP plus carbon additives (LFP/C) with 1.75 ± 0.05 mm diameter were produced following a patented process [53]. Coated LFP powders (1.29% wt carbon, Nanografi, Turkey) were used as electroactive materials, whereas two different carbon additives were evaluated as electronic conductors: glassy carbon (GC) microspheres (Sigradur G HTW, Denmark) and SuperC45 conductive carbon black (CB, IMERYS, Belgium). Different mixtures of LFP and carbon additives were prepared via ball milling for 4 h at 200 rpm using a Planetary Micro Mill Pulverisette 7 (FRITSCH, Denmark).
The filament fabrication begins with the production of the organic binder by mixing dibutyl phthalate, ethylene glycol, and polyvinyl butyral (Merck KGaA, Denmark) in a 1:2:2 ratio. Ceramic powders are then added to this mixture to produce a ceramic barbotine that consolidates rendering a green body, which is extruded at 70 • C using a Filastruder Kit (USA). The so-produced filaments are then dried at 60 • C for several hours.
Several different filaments were manufactured as described in table 1. Two of them, LFP70 and LFP72, did not contain carbon additives to determine the adequate solid load (70 and 72% wt) for good printability conditions. All filaments with carbon additives (CB, LFPXCB, or GC microspheres, LFPXGC, where X is the % wt of the additive) were adjusted to a 70% wt solid load. Filaments free of carbon additives were mostly The nozzle diameter (ND) was 0.4 mm because it is appropriate to the geometry-dimension ratio; printing speed was set to 20 mm·s −1 to obtain a homogeneous distribution of fused filament in the printing surface. The layer height (LH) to ND ratio was set to 0.25, i.e. 100 µm layer thickness to achieve good adherence between layers. Disk 1 geometry was used to optimise the thermal treatments and to perform the electrical characterisation, whereas Disk 2 geometry was employed for the electrochemical tests.
The debinding and sintering processes required thermal treatments up to 500 • C-900 • C for 6 h using 1 • C·min −1 ramp rates and considering three intermediate isothermal stages at 100 • C, 200 • C and 300 • C (for 2, 4 and 6 h, respectively). The debinding process was defined according to thermogravimetric analysis (TGA) using a Jupiter STA 449 C thermobalance (NETZSCH, Denmark) in flowing Ar gas (25 ml·min −1 ) from 50 • C to 500 • C at a ramp rate of 10 • C·min −1 .

Structural and microstructural characterisation
The phase stability of the 3D-printed LFP and LFP/C monoliths was analysed by x-ray diffraction powder (XRD) after thermal treatments under oxidising (air) and inert (dry Ar) conditions. XRD experiments were performed using a PANalytical's X'Pert PRO MRD diffractometer (PANalytical, Netherlands) with monochromatic Cu Kα 1 radiation (λ = 1.5406 Å) operating at 45 kV and 40 mA in the 2θ = 10-80 • range. XRD patterns were analysed with FullProf software [58]. Samples with the highest carbon/ceramic ratio were compared to those without additives and with binder-free monoliths.
Scanning electron microscopy (SEM) images were obtained using a JEOL JSM-6490LV microscope (JEOL, Japan) at an accelerating voltage of 5-30 kV and equipped with secondary and backscattered electron and energy-dispersive x-ray spectroscopy detectors (Oxford Inst., United Kingdom).
High-resolution transmission electron microscopy images were collected in a JEOL 2100 microscope (JEOL, Japan) operating at 200 kV and equipped with an ultra-high resolution (0.19 nm) polar piece and an Orius Gatan 2 × 2 Mpi digital camera (Gatan, USA). Specimens were prepared via suspension of LFP powders in acetone and depositing a drop on a holey-carbon Cu-grid (200 mesh, SPI Supplies).
The relative density and the porosity of the ceramic material was determined with an Ultrapyc 1200e pycnometer (Quantachrome, USA). Additionally, an evaluation of the pore geometry (including diameter, perimeter and circularity) was performed using ImageJ software [59]. Cross-section SEM images (×4000 magnification, 32 × 22 µm 2 ) were segmented based on porosity values obtained via pycnometry considering a minimum pore area of 0.01 µm 2 (areas below this value suppose a negligible fraction, <0.05%).

Electrical and electrochemical characterisation
The electrical conductivity of sintered monoliths (Disk 1) fabricated via FFF with filament free of carbon additives was compared to that of the equivalent disks obtained by conventional press-and-sinter process.
Pristine LFP powders were pressed at 5.2 MPa and sintered at 500 • C for 6 h under dry Ar atmosphere. Next, platinum mesh contacts were placed onto both disk faces. Impedance spectra were collected from 300 • C to 50 • C (with 2 h of stabilisation at each temperature) using a 1470E CellTest System (Solartron Analytical, United Kingdom) with a 25 mV AC voltage in the 1 MHz to 0.01 Hz frequency range.
The electrochemical characterisation of sintered ceramic LFP/C electrodes (Disk 2) was carried out using a Swagelok-type half-cell test configuration [60]. Cells were assembled in an argon glovebox, using LFP/C-based materials and lithium metal as cathode and anode, respectively, both soaked in 1M LiPF6 in EC/DMC 1:1 liquid electrolyte solution (Merck KGaA, Denmark) and grade GF/D filter paper (Whatman plc, United Kingdom) as separator. A BCS-815 Battery Cycling System (Biologic Syst. Corp, France) was used for charge/discharge testing at a constant current density of 85 mA·g −1 (C/2) and 170 mA·g −1 (1C) in the 4.1-1.9 V range. Lithium insertion/deinsertion was limited to x = 0.89, corresponding to 155.5 mAh·g −1 , i.e. the nominal capacity stated by the commercial powder supplier [61]. Composite electrodes with 5 wt% PVDF (Merck KGaA, Denmark), 10 wt% SuperC45 CB and 85 wt% of active material (LFP) were used as control electrodes. The rate capability was also analysed in the C/2-10C range for the printed and sintered LFP/C electrodes.

Filament optimisation and 3D printing
Considering our previous results in the development of LiB electrode materials via FFF [54], LFP filaments were loaded with 70 and 72 wt% of ceramic powder. Nevertheless, as included in table 1, only filaments with 70 wt% of solid load were considered as higher ceramic contents led to problems upon 3D-printing such as nozzle clogging or non-homogeneous flow. In any case, this upper limit implies a remarkable increase compared to previous work on LiB components produced via FFF, i.e. 52 wt% [45].
The next step consisted of the optimisation of the carbon additives (type and composition) included in the filament formulation to further improve the electronic conductivity. Thus, the carbon/ceramic ratio was gradually increased while keeping a 70% of overall solid load according to previous experiments of printability. Finally, high-quality printable filaments were obtained by using 3 and 5 wt% of CB and 10, 20 and 30 wt% of GC as electronic conductor additives in the filament formulation. Those filaments manufactured with 7 wt% of CB and 40 wt% of GC could not be used for high quality printing and hence were discarded.
Particle size and distribution of the ceramic powders in the polymeric matrix are key parameters for the printability of highly loaded filaments [62,63], and indeed large particles or agglomerates may lead to nozzle clogging [54]. In the present case, the distribution of the carbon additives is also decisive, not only to facilitate extrusion but also to ensure good electrical contact at the electrode. Images corresponding to the cross-section of the filaments with CB additive showed 0.5-2 µm LFP spherical particles (in good agreement with the supplier's datasheet [61]) and/or agglomerates evenly distributed in the polymeric matrix (figure 1(a)). No differences in the particle size distribution and morphology of the active material were observed in GC-based filaments (figure 1(b)), although these additive particles (1-6 µm) could be clearly discerned as they are much larger than black carbon nanoparticles.
Regarding printed monoliths, LH/ND ratio was a key parameter to avoid formation of air gaps [64]. Although those gaps would provide accessibility of the electrolyte to the active material, enhancing charge transfer and providing higher rate capabilities [65], they may also have a negative impact on both the mechanical and electrical properties [66][67][68][69]. Figure 1(c) confirms the absence of air gaps in printed monoliths after sintering using the LH/ND ratio of 0.25. Disk 1 and Disk 2 green bodies of LFP-based materials are shown in figure 1(d).

Debinding process and carbon species preservation
As the presence of the polymeric binder used for filament production in the printed electrodes reduces significantly both the electrical conductivity and electrochemical performance, it is necessary to remove it. However, filament formulation includes carbon additives, which may be affected upon binder removal. In order to preserve carbon additives, debinding process is exclusively carried out under flowing Ar.
TGA analysis for LFP70 filament (figure S1(a)) shows a weight loss of ≈29 wt% throughout the entire thermal treatment, mainly in the 200 • C-500 • C temperature range, which may be ascribed to the degradation of the organic components (solvents, binder, plasticisers, etc) [53]. It should be noted that this value is somewhat smaller than the nominal composition, which can be explained because of extra carbon formation due to pyrolysis of the organic compounds in the filament. Such additional carbon has proven beneficial in terms of electronic conductivity and electrochemical performance and explains the use of carbon coated LFP as reference material instead of uncoated powder [70]. Indeed, the decomposition of polymeric additives under inert atmosphere is a widely used method for carbon coating [71]. On the other hand, filaments with carbon additives exhibited analogous behaviour upon thermal treatment (figures S1(b) and (c)), which points out that such carbon additives do not degrade at 500 • C under inert conditions.
The estimation of the active mass is crucial for the electrochemical tests. Based on TGA analysis, table 1 shows the approximate final composition of the printed and sintered monoliths according to the filament used for printing.

Phase stability upon debinding and sintering
The reactivity/stability of the printed components and the formation of secondary phases must be monitored to prevent poor electrochemical performances. In the present case, this is particularly relevant when considering that there are several carbon sources, i.e. organic binder, CB or GC, that may cause decomposition of the active material [72][73][74].
The XRD patterns of the additive-free printed and sintered specimens reveal that LFP is not stable under oxidising atmospheres at 500 • C ( figure 2(a)). The presence of oxygen facilitates the oxidation of Fe 2+ to Fe 3+ , and consequently Li 3 Fe 2 (PO 4 ) 3 and FeO 2 appear as secondary phases [75]. On the contrary, the formation of secondary phases was avoided in printed monoliths without carbon additives under dry Ar conditions ( figure 2(a)). Le Bail fitting revealed an orthorhombic structure, Pnma space group, being the lattice parameters a = 10.379(1), b = 6.036(5) and c = 4.716(2) Å, and V = 295.487(2) Å 3 . Similarly, the XRD patterns of pressed and sintered coated LFP powders (i.e. without binder addition) did not exhibit extra peaks related to the formation of secondary phases using the same thermal treatment ( figure 2(b)); the space group is maintained, with cell parameters a = 10.375 (2) Similarly, printed monoliths containing conductive carbon additives were also evaluated, particularly those with the largest carbon/ceramic ratio i.e. LFP5CB and LFP30GC ( figure 2(b)). None of them showed the presence of secondary phases after thermal treatment under Ar atmospheres, which suggests that the original crystal structure was preserved and justifies the use of non-oxidising conditions during the debinding and sintering thermal treatments.
Moreover, LFP is stable in dry Ar atmosphere up to 850 • C. Above this temperature, the presence of Fe 2 P was detected as a secondary phase (figure 2(a)), as previously reported [76,77]. However, 850 • C cannot be considered straightaway as the most adequate sintering temperature as the carbon coating may degrade upon heating and hence having a negative impact upon the electrochemical performance of the electrodes. The as-received LFP powders exhibit a 5-10 nm thick carbon coating (figure 3(a)), which is partially lost at temperatures as low as 500 • C (figure 3(b)) and therefore resulting in a poorer electrical contact. In the case of the 3D printed electrodes, the carbon coating is preserved after sintering at 500 • C (figure 3(c)), although some uncoated regions appear at higher temperatures ( figure 3(d)). Such improved stability of the carbon coating upon sintering in the 3D printed specimens can be attributed to the organic binder used that prevents faster degradation.

Microstructure
Pycnometry and SEM images processing (table 2, figures 4(a) and (b)) revealed that printed and sintered specimens show a ≈12% lower relative density compared to pressed and sintered counterparts. Such difference is mainly attributed to the closer contact of ceramic particles on uniaxially pressed specimens compared to that of the printed monoliths; furthermore, the organics in the printed samples must be removed, which in turn results in higher porosity [54,78]. Consequently, printed LFP samples after sintering showed larger average pore diameter (1.3 vs. 0.8 µm) and average pore perimeter (6.8 vs. 4.6 µm) than the corresponding pressed specimens, which would improve the electrolyte access and wettability of electrodes and therefore, extending the electrochemical active area [65]. In both cases, pores differ from perfect circles (circularity = 1), showing rather ellipsoidal shapes instead.
A closer look to the pressed specimen reveals that the particle size has barely changed and the grain size and morphology correspond to the initial LFP powders ( figure 4(d)). On the other hand, the 3D printed specimen reveals the presence of particle agglomerates that exhibit necking, i.e. they are in the first stages on  grain growth (figure 4(c)). However, the temperature is not high enough to achieve a higher level of densification. Although full density is not achieved after the sintering process, the so-produced electrodes exhibit enough mechanical strength to ensure correct cell assembly and testing. Figures 4(e) and (f) show a cross section view of the as 3D printed Disk 2 electrode and the same electrode after the debinding and sintering process, revealing a ≈15% shrinkage rate (100 vs. 85 µm).

Electrical conductivity
AC impedance spectroscopy analysis was performed on both 3D-printed (Disk 1) and pressed LFP specimens after sintering. No carbon additives were considered to evaluate the impact of the FFF fabrication process versus conventional uniaxial pressing. The 3D-printed disk exhibits higher conductivity (2.06 × 10 −4 vs. 0.71 × 10 −4 S·cm −1 at 50 • C) in the temperature range evaluated compared to pressed and sintered reference samples ( figure 5). The activation energy is also lower in the 3D printed specimen compared to press and sintered disks using the same LFP raw powder [79]. This could be mostly attributed to the higher carbon content in samples obtained by 3D printing due to pyrolysis of organic compounds/preservation of carbon coating on LFP particles, whilst the pressed and sintered sample may exhibit coating degradation as discussed above ( figure 3(b)). Such degradation does not take place in the reference sample as the sintering temperature was lower, i.e. 400 • C. Indeed, the pressed and sintered specimens tested in this work showed a closer performance to uncoated LFP samples [79]. Therefore, it can be concluded that the presence of the organic binder is beneficial to maintain a good quality carbon coating to ensure electrical contact between LFP particles. These results indicate that FFF manufacturing process does not necessarily imply lower electrical conductivity compared to traditional fabrication processes once the printing parameters and further thermal treatment are properly optimised.

Electrochemical performance
The electrochemical characterisation was carried out on 85 µm thick electrodes (Disk 2, figure 4(f)). The cycling behaviour at C/2 (85 mA·g −1 ) and at 1C (170 mA·g −1 ) of 3D printed electrodes LFPXCB and LFPXGC vs. Li is shown in figures 6(a)-(f) and figures 7(a)-(h), respectively. The aim of these experiments is to assess the electrochemical performance of different electrodes according to the carbon additive (carbon black -CB-or glassy carbon microspheres -GC-) and carbon/ceramic filament ratio used. The average reversible capacities and efficiencies are collected in table 3.
The use of CB as additive in printed electrodes provides a very similar electrochemical profile (figures 6(a)-(d)) to that of the control cell obtained by uniaxial pressing of the composite electrode material. Although the first discharge of conventional LFP:C:PVDF composite electrode yields larger specific capacity than 3D printed electrodes, they stabilise at higher reversible capacity after few cycles, especially in Figure 5. Arrhenius plot of the overall electrical conductivity for LFP disks manufactured by uniaxial pressing ( ) and 3D printing ( ) after sintering, also compared to the bibliography ( , coated LFP, , uncoated LFP) [79]. the case of those fabricated with LFP5CB filament. Both CB-based printed compositions, LFP3CB and LFP5CB, showed a voltage plateau at ≈3.4 V, in good agreement with the reference electrode and previous reports [80]. Nevertheless, variations in the carbon/ceramic ratio have a significant impact on the polarisation, especially during the first charging processes, leading to a considerable difference in the electrochemical performance. Thus, electrodes with higher CB loading showed higher reversible capacity at both cycling rates, 152 mAh·g −1 for LFP5CB and 137 mAh·g −1 for LFP3CB at C/2 rate, i.e. ≈98 and 88% of the maximum theoretical capacity [61]. Similarly, at 1C rate, the reversible capacities were 139 and 121 mAh·g −1 for LFP5CB and LFP3CB respectively (≈90 and 78% of the maximum theoretical capacity). Beyond such minor variations, both electrodes showed high cycling stability and coulombic efficiency at both cycling rates (figures 6(e)-(f)) after 100 cycles, i.e. ≈97% at C/2 rate and ≈98% at 1C. Although the electrodes printed with LFP3CB filament required more cycles to reach the capacity stability (20 vs. 5-10), both compositions maintained coulombic efficiency and reversible capacity after testing, showing reversible capacity losses below 5% upon cycling.
In the case of the electrodes containing glassy microspheres, the results of the electrochemical tests were analogous, with lower capacities during the first cycle, though the LFP20GC and LFP30GC electrodes reached larger reversible capacity values after a few cycles. The higher polarisation observed during first cycle turned into lower performance upon cycling. To overcome such initial limitation, the capacity rate of the experiments was progressively increased from C/5 (34 mA·g −1 ) to C/2 or 1C once a pseudo-plateau is reached during the first cycle. This strategy can be easily appreciated in figures 7(a), (c) and (e) by the irregular profile during the first charge process. It must be noted that this issue was exacerbated when increasing the amount of GC microspheres. Contrary to CB (nanoparticles), glassy microspheres do not distribute so homogeneously around LFP grains due to their geometry and micron-range size. Therefore the microspheres do not provide such a good electrical contact, which causes irregular profiles during the first charge process, particularly at higher C-rates and larger carbon contents. Nevertheless, GC-based electrodes provide fairly similar electrochemical profiles (figures 7(a)-(f)) to those of the reference composite electrode (green dashed line), with an average voltage plateau at ≈3.4 V. Regarding the carbon/ceramic ratio, electrodes printed with the lowest content of GC microspheres (LFP10GC) exhibited the lowest reversible capacity independent of the rate tested (figures 7(g)-(h)), i.e. 142 mAh·g −1 at C/2 rate and 120 mAh·g −1 at 1C rate. Electrodes printed with LFP20GC showed the best performance and stable behaviour, 150 mAh·g −1 at C/2, approximately 96% of the theoretical maximum capacity of the commercial powder [61]. On the other hand, at 1C, the LFP30GC-based electrode yielded the highest specific capacity, i.e. 134 mAh·g −1 , which is ≈86% of the theoretical maximum capacity centred at ≈3.4 V.
The electrodes printed from LFP20GC and LFP30GC filaments showed comparable coulombic efficiencies to CB-based electrodes (95%-98%) at both rates. Nevertheless, GC-based electrodes required more cycles to stabilise the reversible capacity when increasing the carbon/ceramic ratio, e.g. 10-15 cycles for electrodes printed with LFP10GC filament compared to more than 30 cycles for those fabricated with LFP30GC filament.
In any case, all compositions exhibit good capacity and coulombic efficiency retention with comparable losses after 100 cycles to those obtained with CB additives.
Additionally, C-rate capacity tests were carried out for the same filament compositions, from C/2 to 10C rate (figures 8(a) and (b)). For CB-based electrodes, both compositions showed high cycle stability, combined with full recovery of the initial capacity at C/2 even after cycling at rates as high as 10C ( figure 8(a)). Once the stability was reached after the initial cycles, the coulombic efficiency is quite stable except for the first cycle of every current change, especially at high C-rates as previously reported for analogous thick electrodes [81]. Samples printed using LFP5CB filament exhibited higher reversible capacity and coulombic efficiency in all C-rates, as expected according to the cycling performance results above, although both samples show a significant performance decrease at the highest C-rate, i.e. 10C (≈20 mAh·g −1 , 13% of the maximum capacity). At 5C, electrodes manufactured with LFP5CB exhibited ≈80 mAh·g −1 , corresponding to a yield above the 50% of the theoretical commercial capacity [61].
Regarding GC-based electrodes, they also revealed fairly good cycling stability and full recovery of the initial capacity after working at 10 C rate ( figure 8(b)). Coulombic efficiency showed fast stabilisation after the first cycle at fixed C-rate, as occurred for CB-based electrodes. Again, the electrode performance increased with the carbon/ceramic ratio. Indeed, electrodes with GC contents above 10% showed performances analogous to those of CB-based, although their capacity was improved at higher C-rates, i.e.≈60 mAh·g −1 at 10 C for LFP30GC that corresponds to ≈39% of the theoretical commercial capacity, which is significantly higher than that obtained for CB-based electrodes.
It should be noted that, even under demanding conditions (e.g. 1C rate), the electrochemical performance is fairly similar to the values reported for state-of-the-art LFP-based cathodes [80] and significantly higher compared to electrodes obtained by 3D printing [45,46].
In general terms, all the 3D-printed electrodes exhibited good stability and capacity retention, high coulombic efficiency and charge/discharge voltage plateau. Electrodes with the highest carbon/ceramic ratio (LFP5CB and LFP30GC) have proven better electrochemical performance independent of the carbon additive used, although some differences can be highlighted. CB-based printed electrodes needed considerably less cycles to reach capacity stabilisation. Indeed, the increase of CB content accelerates the electrode activation, while an increase in the amount of GC microspheres does not cause the same trend. On the other hand, higher GC contents leads to better results when the current density increases at higher C-rates.
The difference between CB and GC can be attributed to the electrical contact with the particles of the active material. As mentioned above, CB nanoparticles distribute homogeneously around LFP whereas the larger micro-size GC microspheres are not so effective and this results in large polarisations upon charging, which is particularly at larger C-rates. Again, larger C/LFP ratios led to larger capacities in all cases, highlighting that C-additives are required to produce the best electrochemical performances, although in the present case the limit of 5% for CB and 30% GC is imposed by filament printability.
All in all, the highest reversible capacity at C/2, 1C and 2C rates was found for LFP5CB, although as the tests shifted to higher C-rates, GC electrodes perform better and, at 10C rate, LFP30GC would be the best choice. Nevertheless, as previously mentioned, all GC-based electrodes required special starting conditions to avoid large polarisations during the first cycle. Considering all the above, and as good printability was not compatible with further CB content, LFP5CB electrodes can be considered the best option among tested for the fabrication of full ceramic LFP/C cathodes.
As already mentioned, the electrochemical performance of the 3D-printed LFP electrodes in this work is comparable to that obtained for LFP electrodes produced by conventional processes, with no need of polymeric binder, which results in considerably higher energy densities. Moreover, these results exceed those found in the literature for LFP-based electrodes produced via FFF [45], possibly due to the large amount of PLA (40 wt%) and plasticiser required as well as the higher final thickness (200 µm), which explains the extremely high polarisations reported at C/2 rate, and consequently, the null reversible capacity. Thinner PLA-based electrodes (100 µm) with grid-like patterns also exhibited lower reversible capacities (125 and 90 mAh·g −1 at C/2 and 1C, respectively) despite having considered a 10% extra porosity during printing [46]. On the other hand, binder-free thick LFP-based electrodes (up to 300 µm) produced via powder extrusion techniques [32] exhibited reversible capacities of approximately 50 mAh·g −1 at C/2 rate which suggests that the electrode thickness is a key parameter in the electrochemical performance similarly as occurs in conventional fabrication methods [82]. Further work is on-going to produce geometries that may take advantage of the high reversible capacities observed in thin electrodes and the higher energy density associated to thick electrodes. Nevertheless, this work demonstrates that the combination of FFF and the corresponding debinding and sintering processes allows the production of highly conductive LFP-based electrodes that overcome previous performance limitations and highlights the potential of this methodology to produce high performance ceramic-based components.

Conclusions
Novel LFP-based filaments for FFF-3D printing have been developed and used for the production of binder free Li-ion battery electrodes. In addition to LFP as electrochemically active material, GC microspheres and CB were considered in the filament formulation to act as electronic conductive additives to reach a solid content up to 70 wt%. Filaments accepted up to 30 wt% of GC microspheres and 5 wt% of black carbon while keeping good quality for the 3D printing process. After optimisation of both the 3D printing process and the corresponding thermal treatments for debinding and sintering, highly reproducible binder-free electrodes were fabricated with thicknesses down to 85 µm, exhibiting higher levels of porosity compared to electrodes produced via press and sintering.
Debinding and sintering was performed up to 500 • C in Ar atmospheres to ensure phase stability and preserve the carbon additives in the formulation to guarantee good electrical contact between the particles of the active material. Using this strategy, LFP-based 3D printed electrodes exhibited conductivities up to 2.06 × 10 −4 S·cm −1 at 50 • C, which is somewhat higher than the values reported for conventional pressed and sintered specimens [79].
The electrochemical tests performed in half-cell configuration yielded reversible specific capacities as high as 152 or 150 mAh·g −1 at C/2 and coulombic efficiencies of 97.6 or 96.3% after 100 cycles for electrodes fabricated with a 5% of CB (LFP5CB) or a 20% of GC (LFP20GC), respectively. Similarly, reversible specific capacities up to 139 or 134 mAh·g −1 and coulombic efficiencies of 98.0 or 97.7% after 100 cycles at 1C rate have been obtained for LFP5CB or LFP30GC electrodes, respectively. Electrochemical performance was completed by C-rates analysis, showing full recovery of reversible capacity even after 10C rate, with highly valuable specific capacities of ≈80 mAh·g −1 at 5C for LFP5CB and 60 mAh·g −1 at 10C for LFP30GC, above 50% and 39% of the theoretical maximum capacity. These values are fairly similar to the current state-of-the-art of LFP-based cathodes produced via conventional fabrication processes and, more important, significantly improve those previously reported for LFP 3D printed electrodes. Such enhancement of the electrochemical performance is largely due to the lack of non-active materials in the electrodes produced in the present work.
These results demonstrate the potential of FFF-3D printing technology to develop and fabricate high performance LIB electrodes and paves the way to explore free-form geometries to produce a new generation of highly efficient customised electrochemical devices using a low-cost technology.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).