Three-Dimensional Printing of a LiFePO4/Graphite Battery Cell via Fused Deposition Modeling

Among the 3D-printing technologies, fused deposition modeling (FDM) represents a promising route to enable direct incorporation of the battery within the final 3D object. Here, the preparation and characterization of lithium iron phosphate/polylactic acid (LFP/PLA) and SiO2/PLA 3D-printable filaments, specifically conceived respectively as positive electrode and separator in a lithium-ion battery is reported. By means of plasticizer addition, the active material loading within the positive electrode is raised as high as possible (up to 52 wt.%) while still providing enough flexibility to the filament to be printed. A thorough analysis is performed to determine the thermal, electrical and electrochemical effect of carbon black as conductive additive in the positive electrode and the electrolyte uptake impact of ceramic additives in the separator. Considering both optimized filaments composition and using our previously reported graphite/PLA filament for the negative electrode, assembled and “printed in one-shot” complete LFP/Graphite battery cells are 3D-printed and characterized. Taking advantage of the new design capabilities conferred by 3D-printing, separator patterns and infill density are discussed with a view to enhance the liquid electrolyte impregnation and avoid short-circuits.

such as electrodes, separator, solid polymer electrolyte or also current collector can be customized, thus enabling the direct incorporation of microbatteries and electronics within the final optimized design 3D object [26][27][28][29][30] .
On the other hand, to reach enhanced electrochemical behavior, 3D battery designs were introduced 31,32 . 3D architecture allowing two-dimensional or three-dimensional diffusion of the lithium cations were disclosed, thus increasing the electrochemical active surface present on the same footprint area, power and specific capacity 33 . This was confirmed by Ragones et al. 25 who reported the modeling of interlaced electrode 3D networks and evaluated the area gain of those architectures. More recently, an outstanding work done by Trembacki et al. 34 demonstrated through simulations that 3D battery designs (gyroid and Schwarz P) perform significantly better than the particle bed geometry (2D) with energy density improvements of 3.7x − 6.9x observed at the highest power density simulated (12C).
Several groups 13,35,36 already demonstrated the preparation of 3D-electrodes separately (positive or negative). However, as a consequence of both 3D electrodes surface roughness, the mounting of the full 3D device was described as particularly challenging with a view to prevent any short-circuit. Positive and negative electrodes interpenetration concerns may be overcome by using additive manufacturing technologies (AM). Within this framework, an escalation of investigations concerning the AM technologies, applied to energy storage field were reported lately 7,19,25,[37][38][39][40][41][42][43][44][45][46][47][48] . Among them, liquid deposition modeling (LDM) and fused deposition modeling (FDM) were seriously investigated recently in order to 3D-print electrochemical storage systems such as lithium-ion batteries.
LDM is a method in which an ink employed as material source for the 3D-printer is extruded from a syringe while the latter is moved across a platform controlled by computer. Thin layers of material are deposited in order to build the desired 3D object in many successive passes. Regarding this process (Table 1), pioneering work was reported by Sun et al. 43 who described the first 3D-printed Li-ion microbattery. They prepared cellulose-based inks of Li 4 Ti 5 O 12 (LTO) and LiFePO 4 (LFP) nanoparticles suspended in an aqueous solution for the formulation of negative and positive electrodes. Those inks were 3D-printed through a syringe, dried and sintered at 600 °C for 2 h in an inert atmosphere. Discharge electrochemical measurements for half-cell of LFP led to specific capacity of 160 mAh g −1 of active material at current density of 170 mA g −1 (1 C), corresponding to areal capacity of 1.6 mAh cm −2 at 1.7 mA cm −2 . In the other hand, half-cell of LTO displayed a specific capacity of 131 mAh g −1 of active material at current density of 175 mA g −1 (1 C), equals to areal capacity of 1.4 mAh cm −2 at 1.87 mA cm −2 .
In order to obtain the required viscosity enabling 3D printing but also to improve the electrode's electrical conductivity, Fu et al. 37 used aqueous inks of LTO/graphene oxide (GO) and LFP/GO to feed the printer. Both printed electrodes were subsequently freeze-dried to eliminate water solvent and solidify the 3D structures. A sintering procedure was applied under Ar/H 2 (600 °C for 2 h) to allow the formation of reduced graphene oxide (rGO). The finally obtained LFP/rGO and LTO/rGO half cells were electrochemically tested using a polyvinylidene fluoride-co-hexafluoropropylene (PVDF-co-HFP) with Al 2 O 3 3D-printed separator. At current density of 10 mA g −1 (C/17), specific capacities of 164 and 185 mAh g −1 of active material were respectively demonstrated. These electrodes were then assembled into a full battery depicting a capacity about 100 mAh g −1 during ten cycles at a specific current of 50 mA g −1 .
Li et al. 49 reported the 3D-printing of a LiMn 2 O 4 /carbon black/PVDF ink (wt.% LMO/CB/PVDF 85.5/6.5/8) dispersed in N-methyl-2-pyrrolidone solvent (NMP) to fabricate electrodes. After printing, a voltage of 10 kV was applied at a distance of 1.25 cm for 10 minutes. A hot plate (at 120 °C) was finally used to remove solutions quickly. In the proposed processing, the applied electric field controls the microstructures of manganese-based electrodes, while additive manufacturing controls macro-3D structures. As compared to a conventional laminated structure (1.8 mAh.cm −2 at C/10), the macro-micro controlled structure showed improved performances (3.5 mAh.cm −2 ).
As exemplified hereabove 37 , 3D-printed separator loaded with various ceramic fillers were reported in literature. Furthermore, Liu et al. 50 prepared a PVDF-co-HFP/boron-nitride (BN) ink using dimethyl formamide as solvent to 3D-print separator. After printing, it was dried directly at room temperature. Authors showed that BN-separator dispenses quick heat dispersion and a homogeneous thermal distribution at the interface during the Li plating process increasing electrochemical performance of the lithium-ion battery.
Finally, an important milestone was reached by Wei et al. 51  The cathode, anode, separator, and packaging inks were housed in separate 3 mL syringes. 3D-printing was performed by using an Ar-powered fluid dispenser to pressurize the barrel up to 700 psi to control the flow rate. After UV-curing, these fully 3D printed and packaged LIBs, which were encased between two glassy carbon current collectors, deliver an aerial capacity of 4.45 mAh cm −2 (equivalent to 17.3 mAh cm −3 ) at a current density of 0.14 mA cm −2 .
To summarize, while significant advantage of LDM versus FDM is the possibility to achieve high active mass loading, main drawback of the LDM is certainly the requirement of performing post-processes such as freeze-drying and sintering prior to its use 37,43 . Furthermore, ink viscosity is a key factor to consider as it restricts the height of the successively 3D-printed object. Hence, many parameters impacting this important point must be studied thoroughly such as the introduction of surfactant and/or the active material granulometry among others.
FDM technology can overcome the latter concerns, nevertheless, only thermoplastic polymers such as polylactic acid (PLA), polyethylene terephthalate glycol (PETG), acrylonitrile-butadiene-styrene (ABS), polycarbonate (PC), or also polypropylene (PP) amongst the most employed, can be used to prepare a filament suitable for an FDM 3D-printer.
FDM produces cheap complicated items with nearly no material leftover 20,23,52 . This technology consists of depositing, through a heated nozzle of which the movement (XYZ-axis) is meticulously computer-piloted, thin layers of melted thermoplastic material in order to build the desired 3D object in many successive passes. As the thermoplastic material is heated only few degrees above its melting temperature to be deposited, it instantaneously cools down and solidifies after extrusion. Recent commercial FDM 3D-printers layer thickness resolution ordinarily reaches 200 µm for the first layer and can be commonly enhanced up to 50 µm for the successive layers. For higher resolution purposes, layer thickness resolution can be improved by customizing the machine and adjusting computer program settings 53 . FDM is now employed for fast prototyping in a wide variety of purposes comprising aerospace, automotive but also medical fields.
Regarding the FDM technique focused on lithium-ion batteries (Table 2), pioneering work was reported by Foster et al. 44 in 2017. They described the fabrication of a 1 mm thick 3D-printed negative electrode disc by operating a commercial graphene-based polylactic acid filament (graphene/PLA) as material source. Because of the little amount of active material (only 8 wt.% graphene compared to 92 wt.% PLA) within the filament (103 mg of active material per cm 3 of composite), low discharge specific capacities of 15.8 mAh g −1 of active material (1.26 mAh g −1 of the total composite equivalent to 1.63 mAh cm −3 considering the electrode volume) were reached at current densities of 10 mA g −1 (C/37).
Reyes et al. 19 reported recently the 3D printing of a complete lithium-ion battery (LTO-LMO full cell) by FDM for the first time. This was possible through the formulation of PLA/LTO/conductive additives and PLA/ LMO/conductive additives filaments employed to feed the FDM 3D-printer (6 vol% for LTO anode and 4 vol% for LMO cathode). Commercial filament of PLA was used as separator and infused in 1 M LiClO 4 in 1: 1 vol% ethyl methyl carbonate: propylene carbonate (EMC: PC). Fully 3D printed assembled coin cell, as well as "one-shot" 3D printed devices and wearable electronic devices with integrated batteries were described. The volumetric capacity of the full-cell was reported to be equals to 0.25 mAh cm −3 while negative and positive half-cells displayed respectively 0.34 and 0.71 mAh cm −3 at a current density of 10 mA g −1 .
Unfortunately, due to the printability requirements for the FDM process, active material loading was kept relatively low in those last studies thus impacting severely the electrochemical performances 19,25,44 . However, as reported in our previous work 42 , this limitation can be overcome by introducing a plasticizer. Indeed, we reported a highly loaded 3D printable graphite/PLA filament specifically conceived to be employed as negative electrode in a lithium-ion battery and to feed a conventional FDM 3D printer. Active material content (graphite) within the filament, was increased as high as possible (49.2 wt% of graphite in the total composite thus corresponding  www.nature.com/scientificreports www.nature.com/scientificreports/ to 773 mg of active material per cm 3 ) to improve the electrochemical performances while preserving enough mechanical strength to be printed thanks particularly to the addition of poly(ethylene glycol) dimethyl ether average Mn ∼ 500 acting as plasticizer. Hence, an unprecedent reversible capacity for a negative electrode disc obtained via FDM was achieved: 200 mAh g −1 of active material (99 mAh g −1 of the total composite or also to 154.6 mAh cm −3 ) at current density of 18.6 mA g −1 (C/20) after 6 cycles and 140 mAh g −1 of active material (69 mAh g −1 of the total composite or also 108.2 mAh cm −3 ) at current density of 37.3 mA g −1 (C/10).
In order to print the complete lithium-ion battery through FDM, filament formulation of the positive electrode and separator is now required. Here, this work was focused on the development and optimization of LFP-PLA and PLA-SiO 2 composite-based 3D-printing filaments respectively. Through the formulation process, the effect of carbon black as conductive additive in the positive electrode and the impact of ceramic additives in the separator on the ionic conductivity and swelling were investigated. Furthermore, exploiting the new design/ geometry capabilities arising from 3D-printing, separator patterns were considered with a view to enhance the liquid electrolyte uptake. By using both optimized filaments as well as our previously reported graphite/PLA filament for the negative electrode, assembled and "printed in one-shot" complete LFP/Graphite battery cells are 3D-printed and characterized. Assembled cell was mounted after printing separately the different components, while a "printed in one-shot" battery was obtained by pausing the machine and changing the filament between each layer. After each formulation steps (film, filament and/or 3D-printed disc), composite electrodes were characterized thoroughly by performing thermal (DSC), mechanical (tensile test), morphological (SEM), but also electrical and electrochemical analyses. Well aware of the limitations induced by nominal resolution of 3D-printer, this work serves here as proof of concept.

Results and Discussions
The complete formulation route is exhibited in Fig. 1a. As described in our previous work 42 , the solvent method was followed to guarantee a good homogeneity at laboratory scale. Dichloromethane (DCM) was employed as it allows a quick dissolution of PLA and its relatively low boiling point leads to a rapid evaporation of the solvent after slurry doctor blading. Solvent selection is a crucial stage as it allows the active material and other additives to be evenly accommodated within the polymer matrix.
Slurry formulation for the positive electrode at increasing carbon black (CSP) concentration (Table 3) was prepared by dissolving the PLA pellets into DCM, with a weight ratio PLA/DCM 1:10, under magnetic stirring and at room temperature. Once complete dissolution of PLA was achieved, PEGDME500 acting here as plasticizer was added (40 wt.% ratio PEGDME500/PLA) and mixture was magnetically stirred for 30 minutes. In the meantime, the desired amount of active material and conductive additives were pre-mixed in a mortar before incorporation within the slurry in order to guarantee meticulous mixing (wt.% ratio PLA/AM 40:60). Again, magnetic stirring was performed during 5 h. Finally, homogeneous mixture slurries were spread onto a glass substrate following the doctor-cast technique. Self-standing composite films were achieved at room temperature after complete evaporation of the solvent overnight. Same procedure was followed for the preparation of the separator films (Table 4). After dissolution of the PLA, plasticizer was added (40 wt.% ratio PEGDME500/PLA). Then, amorphous SiO 2 nanoparticles (average size 7 nm), display in Fig. S1, were added in the slurry and mixed until a homogeneous mixture was achieved. Subsequently, a free-standing separator film was obtained through tape-casting. Figure 1b exhibits DSC plots of the composite films with various compositions of polymer, LFP, plasticizer and conductive additives. Serving here as a reference, pure PLA displays an endothermal peak related to its melting (Tm) at 146 °C and a well-defined glass transition temperature (Tg) at 63 °C. By adding LFP as active material to the PLA matrix, the endothermal peak related to the melting temperature of the film called PLA/LFP (wt% 40:60) was a little altered to lower temperature emerging at 142 °C and the Tg is not perceptible anymore. Water desorption was detected around 100 °C by the presence of a minor endothermal peak. Because of the important brittleness of the obtained composite film, a plasticizer, poly(ethylene glycol) dimethyl ether average Mn~500 (PEGDME500), was incorporated as discussed in our previous study 42 . Compared to PLA/LFP sample which only shows a clear Tm peak, a little exothermal crystallization peak (Tc) around 80 °C is induced through the incorporation of a plasticizer, in good agreement with literature 54 . This is emblematic of plasticized thermoplastics, as plasticizers may facilitate crystallinity by enhancing chain mobility. As the temperatures difference Tc-Tg is reported to be constant (~80 °C) in the whole range of plasticizer concentration 54 , the glass transition temperature (Tg) of the PLA polymer matrix should also appear in the range of the first endothermal peak at 4 °C imputed to the plasticizer PEGDME500 fusion. Plasticized PLA/LFP films display Tm inferior than without plasticizer, diminishing from 142 °C to about 132 °C.
In parallel, in order to enhance the electrical conductivity within the positive electrode, samples containing different amount of CSP (weight ratio LFP:CSP equals to 100:x with x from 5 to 20) were prepared. By increasing the CSP content, endothermal peaks at 4 °C and 132 °C, respectively corresponding to the PEGDME500 fusion and the composite film melting, remain unchanged. This behavior is however different for the exothermal crystallization peak (Tc) previously observed at 80 °C (0%CSP sample). Indeed, through the addition of CSP, crystallization peak is slightly altered to lower temperature, reaching 74 °C for the 20%CSP sample. This behavior may be explicated by the CSP presence within the PLA matrix impacting crystallinity behavior at the microscopic scale.
Extrusion and 3D-printing were concentrated on the most electrochemically-encouraging 10%CSP film sample hereafter detailed. Plots achieved from DSC for the analogous film, filament and 3D-printed disc (Fig. 1c), display similar Tm value equals to 132 °C. Nevertheless, endothermal wide peak intensity related to the PEGMDE500 fusion at 4 °C tends to drop following extrusion and 3D-printing (performed respectively at 170 and 195 °C), indicating partial plasticizer evaporation occurs during both stages. Conveniently, the lower plasticizer content still confers sufficient filament flexibility in order to feed the 3D-printer. Mechanical performances such as the Young's modulus (223 ± 20 MPa), tensile strength (7.36 ± 0.45 MPa) and elongation at break (14.04 ± 1.35%), were determined from the stress-strain curves. As expected, those values are very low (except elongation at break) compared  55 due to the high loading of composite material here. Furthermore, after 3D-printing, the exothermal peak of crystallization area emerging at 75 °C is more intense. This behavior may be explained by the fast cooling allowed by the fans, thus leading to a less crystallized state of the polymer chains.   www.nature.com/scientificreports www.nature.com/scientificreports/ From our practical experiences, with a view to obtain enough mechanical performances to be printed, the total amount of charges (sum of active material and conductive additives) within the filament must not exceed a total of about 50 vol% when a plasticizer such as PEGDME500 is introduced. On the contrary, without the plasticizer introduction, the total amount of charges must not exceed about 30% otherwise the PLA-based filament will not be printable. As displayed in Table 3, in this study, the most-promising 10%CSP positive electrode filament contains 35 vol% of LFP and 4 vol% of CSP, which represents a total volume of charges equals to 39 vol% of the total composite. A compromise must be found with a view to enhance the electrochemical performances (through a high amount of charges) while still maintaining suitable mechanical properties (thanks to polymer matrix and plasticizer).
Porosities for the 10%CSP positive electrode after each step of the formulation (film, filament and 3D-printed disc) were calculated considering the complete plasticizer evaporation. While the film obtained after tape-casting originally depicts 17% porosity, it can be seen that after extrusion the filament displays only 14% of porosity. This may be easily explained by the extrusion process itself compacting the material. The 3D-printed disc displays about 29% porosity which is explicated by the successive pathways taken by the nozzle leading to a surface inhomogeneity creating high macro-porosity.
The effect on electronic conductivity of carbon black, CSP, was studied. For samples obtained after slurry casting and labeled in Table 3 (weight ratio LFP:CSP equals to 100:x with x from 5 to 20), specific impedance spectra were investigated at temperatures ranging between 20 °C to 50 °C. The achieved Nyquist and Bode curves are in accordance with an electronic conductor classical behavior as Nyquist plot portrays only Z' real part while |Z| magnitude is remained steady for all frequencies. The electrical conductivity, for all samples of equal thickness and area, is rising with temperature as exhibited in Fig. 2a and activation energy values varying between 0.0001 eV and 0.052 eV were estimated. Panabière et al. 56 showed that electrons transport mechanism in carbon/polymer mixtures used in battery field was tunneling with no temperature dependence (metal-like behavior). Nonetheless, such temperature dependence would be perceived when strong dipoles are adsorbed at the carbon particles surface. In this respect, it is not surprising to observe activation energy values increasing with higher CSP content. Conductive additive contributes to greater conductivity values compared to the 0%CSP sample without additives, which is in good agreement with what was reported by Guy et al. 57 However, here, a higher loading of CSP is needed to reach similar values 57 , which is undoubtedly due to the high quantity (32.8 wt%) of the polymer matrix (PLA) having a detrimental effect on the homogenization and CSP distribution which may be isolated. Amongst the various film samples tested, 20%CSP exhibits the greatest values of electrical conductivities. Nevertheless, it is worth noticing that conductivity values are still relatively low due to the LFP insulator behavior (enhanced up to 0.16 S.cm −1 for the 20%CSP sample at 20 °C) as compared to values obtained in our previous study 42 for the graphite-based negative electrode (0.40 S.cm −1 at 20 °C). Carbon coating is therefore important when LFP is considered as active material 58 . Here, the incorporation of carbon additives such as CSP is essential when considering exotic composite compositions with such an important amount of PLA polymer matrix.
On the other hand, a thorough characterization study was performed to establish the electrochemical behavior of the two most electrically conductive film samples 10%CSP and 20%CSP (Figs. 2b and S2). Note that commercial glass fiber was used as separator for the following experiments. For each sample of the same size (60 μm-thick and 11 mm diameter, surface equal to 0.950 cm 2 ) acquired after doctor blading (described in Table 3), the potential profiles versus specific capacity based on the active material and specific capacity versus cycle number were examined at diverse current densities (8.5, 17, 34, and 85 mA g −1 of active material equivalent to C/20, C/10, C/5, and C/2). Among those samples, it was shown that 10%CSP film displayed the best electrochemical performances reaching specific capacity value up to 165 and 162 mAh.g −1 of active material at C/20 and C/10 respectively. It appears that 20%CSP film displays slightly inferior capacity values. Besides, this latter composition was discarded as CSP tends to agglomerate thus affecting the sample homogeneity. As theoretical capacity was almost reached with the 10%CSP sample and as no homogeneity issues were observed, it was consequently decided to focus our studies on this optimized sample for the positive electrode. Consequently, the corresponding 10%CSP filament (756 mg of active material per cm 3 of composite) was produced and 200µm-thick 3D-printed discs were characterized electrochemically (Fig. 2b,c). Compared to specific capacity values obtained for the latter 10%CSP 60µm-thick film, values displayed for the corresponding 200µm-thick 3D-printed disc were observed to be much lower: 87 mAh g −1 of active material at C/20 (equals to 43 mAh g −1 of the total composite or also 66 mAh cm −3 considering the total volume of the electrode) and 45 mAh g −1 of active material at C/10 (representing 22 mAh g −1 of the total composite or also 34 mAh cm −3 ). This important gap is due to the difference of thickness between the film (60 µm) and the 3D-printed disc (200 µm) induced by the 3D-printer low thickness resolution on the first layer. 3D-printed disc shows an irreversible capacity during the first cycle of 38 mAh g −1 of active material, equivalent to a percentage loss of 42% at a current density of 8.5 mA g −1 (C/20).  www.nature.com/scientificreports www.nature.com/scientificreports/ Thereafter, it was decided to check the electrochemical performances of the assembled complete system by using both 200µm-thick optimized 3D-printed electrodes. Hence, sample 10%CSP aforementioned, containing LFP as active material, was used as positive electrode while the 3D-printed graphite electrode optimized in our previous study 42 was used as negative electrode. In this latter, as it can be seen in Fig. S3 depicting the microstructure of both 3D-printed electrodes observed by SEM, the CSP particles are evenly dispersed within the polymer matrix. Here, the homemade PLA/SiO 2 separator was used instead of the commercial fiber glass mentioned earlier. It was subjected to ionic conductivity and swelling measurements after electrolyte soaking with various amount of silica.
Indeed, introduction of ceramic fillers within the polymer matrix has attracted substantial attention in literature 59 , due to their ability in improving thermal stability, mechanical strength as well as ionic conductivity of the separators. In order to enhance those last parameters, the nanosilica content was optimized from 0 to 22 wt.%. For each separator sample described in Table 4, specific impedance spectra were investigated at 25 °C under argon after immersion within the electrolyte during 1 h and 10 h (Fig. 2d). First of all, by introducing SiO 2 nanoparticles within the PLA polymer matrix, it was shown that swelling percentage tends to decrease. Indeed, after being immersed 10 h within the electrolyte, 22%SiO 2 , 18%SiO 2 , 13%SiO 2 , 7%SiO 2 and 0%SiO 2 samples depicted a swelling percentage equals to +2%, +4%, +5%, +8% and +14% respectively. Moreover, after 1 h immersed within the electrolyte, conductivity was improved up to 3.96 × 10 −5 S.cm −1 for the 13%SiO 2 sample compared to value reaching only 2.41 × 10 −5 S.cm −1 for sample 0%SiO 2 . After 10 h immersion, the most conductive sample was 7%SiO 2 displaying 1.20 × 10 −4 S.cm −1 , thus corresponding to seven times the conductivity value obtained for the reference 0%SiO 2 (only 1.67 × 10 −5 S.cm −1 ). An explanation of this behavior is that composite separators have higher electrolyte uptake due to the SiO 2 particles introduction improving wettability compared to the pristine separator 60 . Considering the complete plasticizer evaporation, the optimized sample "7%SiO 2 " presents 11.3 mol% of SiO 2 (9.6 wt%), close to the optimized silica amount of 15.7 mol% (4 wt%) reported by Caillon-Caravanier et al. 61 Through the SiO 2 addition, a more porous structure is formed, thus contributing to faster electrolyte uptake. As also described by Caillon-Caravanier et al. 61 , it is worth mentioning that a too high content of SiO 2 can have a detrimental effect on ionic conductivity as it is clearly visible for the 22%SiO 2 sample (8.71 × 10 −6 S.cm −1 after 10 h). When maximum porosity is reached, further addition reduces absorption ability and conductivity by limiting the porous volume and by interaction of some fraction of the lithium cation with silica 62 . Thus, it was decided to use the optimized 7%SiO 2 sample as separator for the following experiments. www.nature.com/scientificreports www.nature.com/scientificreports/ With a view to favor the liquid electrolyte uptake within the system, we also decided to take advantage of the new design capabilities conferred by 3D-printing. Indeed, various separator infill patterns (Fig. 3a) can be easily obtained (automatically generated) by using classic 3D-printing slicer software. One can easily understand that open-designs such as the Hilbert curves, Archimedean chords or Octogram spiral will favor the spread of liquid electrolyte compared to closed-designs (rectilinear, grid, triangle, stars, cubic, line, concentric, honeycomb and gyroid). Moreover, in order to avoid short circuit, next step is to set the infill density properly (measured in percent). A print with low infill density (Fig. 3b) for the separator will favor the electrolyte uptake while it will also favor short-circuits. Thus, a compromise must be reached.
Regarding the assembled complete system, Archimedean chords was chosen among the open-designs as it is the only pattern that can be printed in one fragment and thus be manipulated. Furthermore, infill density was set to 70% as it seems enough to avoid short-circuits. The capacity retentions at 4.25 mA g −1 (C/40) after 1 h impregnation were plotted. As reference, capacity retention for the system using 100% infill density separator disc is displayed in Fig. 3c. Good capacity retention is observed after 1 h impregnation with the 70% Archimedean chords pattern for the separator as it enables faster liquid electrolyte uptake (Fig. 3d). On the contrary, sample with 100% infill density separator disc, displays increasing capacity after 1 h during the first cycles. It is worth mentioning that a very low reversible capacity value of about 15 mAh g −1 of active material was obtained during 10 cycles at 4.25 mA g −1 (C/40). Those poor performances are induced by the 3D-printer first layer thickness limitation as it is not possible to obtain <200 µm electrodes and separator. Thus, balancing of the cell was not possible.
Finally, as reported for the first time by Reyes et al. 19 , we confirmed the printability of the complete battery in a single print that we shall refer to as "one-shot". Complete lithium-ion battery of any shape can be easily printed such as disc, square, our laboratory logo or even a fancy cat (Fig. 4a,b). This was technically possible by a succession of numerous important steps which are described in more details in the experimental part. Filaments were loaded one after another by pausing the printer. Cleaning steps of the nozzle were unavoidable to ensure good homogeneity. Note that each layer was perpendicularly printed to the precedent to favor adhesion. These tedious steps might be simplified in the future by using a multi-nozzle 3D-printer. However, this solution also comes with drawbacks as nowadays the x-y and z calibrations are still difficult and retraction does not always work properly. Filament leftovers can get stuck on the print when the extruder moves and thus leading to eventual short-circuits. www.nature.com/scientificreports www.nature.com/scientificreports/ Here again, in order to favor the liquid electrolyte uptake within the "printed in one-shot" system, we also decided to set an infill pattern for the separator (Fig. 4c). Among the open-designs, Hilbert curves seems to be the most promising as this is the only pattern enabling liquid electrolyte to enter using different paths. Indeed, Archimedean chords and Octogram spiral display only one entry for the electrolyte. Short-circuits are more complicated to avoid in the "one-shot" configuration as printing is performed at 195 °C with this specific 3D-printer to prevent under-extrusion (when not enough material from the filament is extruded during a print, resulting in gaps or missing layers that strongly prejudice the print quality). Indeed, even with a 70% infill density separator (150 µm thick), the above layer corresponding to the positive electrode have tendency to collapse naturally by gravity to fill the gaps created by the separator pattern, thus leading to short-circuits. This phenomenon is easily observable by scanning electron microscopy (Fig. 4d,e).
To prevent this behavior, it was decided to add a 100% infill density separator layer (100 µm) above the 70% infill density Hilbert curves pattern (50 µm) as displayed in Fig. 5a. Narrow caves created by the separator pattern can be observed by SEM/EDS (Fig. 5b). Ceiling of those cavities is homogeneously covered by a separator PLA-SiO 2 thin layer thus preventing any contact between both electrodes (Fig. S4). Noteworthy that this uppermost separator layer might also prevent lithium dendrite propagation while subjected to improper cycling conditions as low temperature and high rate. Figure 5c depicts the capacity retention plots at different C-rates for the one-shot fully 3D-printed disc system. In this configuration, as separator thickness diminution as well as balancing of the cell are now possible, slightly higher reversible capacity values were reached (30 mAh g −1 of active material at C/40, corresponding to 15 mAh g −1 of the total composite or also 6.5 mAh cm −3 considering both electrodes and separator total volume) compared to the assembled cell (15 mAh g −1 of active material at C/40). After cycling, it is worth mentioning that cracks were observed within the negative electrode layer (Fig. 5d). It is well known that the volume expansion ΔV/V is equal to the trace of the strain tensor ∆ = = .ε V/V Trace ( ) 3 xx  with ε xx the extensional strain along the coordinate axe x equal the two others under the assumption of isotropic swelling 63 . Knowing that graphite exhibits a volume (c) Scheme of the 3D-printed in "one-shot" lithium-ion battery using Hilbert curves pattern 70% infill density as separator layer (150 µm thick). (d,e) Backscattered electron SEM images of the short-circuits observed within this system. It is important to note that 7%SiO 2 and 10%CSP sample were respectively used here as separator and positive electrode while our previously reported (cf. ref. 42 ) Graphite/PLA/CSP/PEGDME500 filament was used to print the negative electrode. , higher than the elongation at break closed to 3% obtained from the stress-strain curve of the negative electrode. On the contrary, as expected, this phenomenon is not observed for the positive electrode as it starts to shrink because of the lithium departure upon first charge.
With a view to improve the overall system, next step would be to develop a current collector filament in order to print the first layer subjected to the 200 µm thickness limitation. Hence, it would be possible to print thinner (down to 50 µm) and balanced electrodes and thus further improve the electrochemical performances. On the other hand, graphite granulometry could be decreased to favor the particles dispersion within the polymer matrix, better accommodate the volume constraints and thus prevent cracks upon cycling.

Conclusion
3D-printable PLA/LFP and PLA/SiO 2 filaments, specifically conceived to be used respectively as positive electrode and separator in a Li-ion battery were produced to feed an FDM 3D printer. As reported in our previous study 42 , thanks to plasticizer PEGDME500 addition, the active material content within the positive electrode filament was increased as high as possible to improve the electrochemical performances while still providing enough mechanical properties to be printed. Films containing CSP showed higher electrical conductivity and specific capacity, indicating their addition enables several isolated active material particles to be electronically connected to the percolating network. Because theoretical capacity was almost reached with the optimized 10%CSP film sample, its corresponding filament was employed as material source for the 3D-printer. Furthermore, the impact of ceramic additives (SiO 2 ) on ionic conductivity in the separator was studied. Samples with various amount of silica were subjected to ionic conductivity and swelling measurements after electrolyte soaking. After 10 h impregnation, the most conductive sample was 7%SiO 2 .
For the first time, considering both optimized filaments composition and using our previously reported PLA/ graphite filament 42 for the negative electrode, assembled and printed in "one-shot" complete LFP/graphite battery cells were 3D-printed by means of FDM. Figure 5. (a) Scheme of the 3D-printed in "one-shot" lithium-ion battery using Hilbert curves pattern 70% infill density as first separator layer (50 µm thick) and 100% infill density (100 µm thick) as second separator layer; (b) Backscattered electron SEM image of the system; (c) Capacity retention plots at different C-rate for the 3D-printed in "one shot battery"; (d) Backscattered electron SEM image of the cracks observed within the negative electrode after cycling. It is important to note that 7%SiO 2 and 10%CSP sample were respectively used here as separator and positive electrode while our previously reported (cf. ref. 42 ) Graphite/PLA/CSP/ PEGDME500 filament was used to print the negative electrode. www.nature.com/scientificreports www.nature.com/scientificreports/ Taking advantage of the new design capabilities conferred by 3D-printing, separator patterns and infill density were considered with a view to enhance the liquid electrolyte impregnation and avoid short-circuits. Archimedean chords and Hilbert curves appear to be the most promising patterns to be used respectively within the assembled and the "one-shot" configurations. Infill density values >70% would be studied in more details in future with a view to prevent short-circuits. While low reversible capacity values were reached for the "one shot" system (30 mAh g −1 of active material at C/40, corresponding to 15 mAh g −1 of the total composite or also 6.5 mAh cm −3 considering both electrodes and separator total volume), it is important to understand that separator thickness diminution as well as balancing of the cell were now possible in this configuration, thus improving the performances compared to the assembled cell (15 mAh g −1 of active material at C/40).
This study, by merging both battery and 3D-printing technologies, addressed numerous electrochemical (thickness, electronic and ionic conductivity, electrolyte uptake) and 3D-printing parameters (infill density, infill pattern, perimeters, over and under-extrusion, retraction), and opens the way for a better performing 3D-printed lithium-ion battery.
Finally, as this work acts here as proof of concept, authors are well-aware that for now electrodes and separator patterns are 2D and thus achievable using non-3D printing techniques. Future work, however, would be concentrated on complex 3D-battery architectures that necessitates significant system adjustments and a thorough design optimization. Upcoming research may also be dedicated to mechanically ameliorate the FDM 3D-printer resolution and simplify the tedious steps to print the full-battery in one-shot by using a multi-nozzle configuration.
Filaments formulation. An extruder Filabot Original provided by Filabot Triex LLC, USA was fed consistently with 3 mm × 3 mm composite film fragments in order to obtain classical 1.75 mm diameter 3D-printing filaments. Extruder temperature was set about 35 °C higher than the melting temperature of the composite film determined from DSC. The filament released from the extruder nozzle was subsequently rolled around a spool by using a Filabot spooler (Filabot Triex LLC, USA). The extruder was purged thoroughly with pure PLA before extrusion of each sample. Electrode filaments were subsequently kept in appropriate storage conditions under low temperature to prevent plasticizer evaporation and in confined environment with low humidity. Well aware of the plasticizer evaporation upon time, they were almost immediately (one week of storage maximum) used as material source for the 3D-printer.
Printing. 11 mm diameter separator, positive and negative electrode discs (200 µm thick) and elaborated 3D structures were printed by using a Prusa MK3 3D-printer (Prusa Research, Czech Republic). Nozzle standard input and output diameters of respectively 1.75 mm and 0.4 mm were used. Best resolution in the Z direction is 0.20 mm for the first layer and 0.05 mm for the following. Nozzle temperature was set with this specific printer to 60 °C above the Tm of the composite filament deduced from DSC as fan settings was set to 100%. Bed temperature was set to 60 °C in order to enhance the adherence of the first printed layer. The external vertical shells perimeter setting was set to 0 to favor the electrolyte spread. Prior printing of each sample, nozzle was cleaned thoroughly by printing a 2 cm 3 purge cube with the corresponding filament. The printability of the complete battery in a single print ("one-shot") was technically possible by a succession of numerous important steps: print the first 200 µm layer with the optimized negative electrode filament, put the printer in pause, remove the negative electrode filament, load the separator filament, ensure it is well loaded, make sure nozzle is cleaned with the separator filament, print layers of separator above the precedent negative electrode, put the printer in pause, remove the separator filament, load the optimized positive electrode filament (10%CSP), ensure it is well loaded, make sure nozzle is cleaned with the corresponding filament, print layers of positive electrode above the precedent separator layer. Differential scanning calorimetry (DSC). Thermal studies of films, filaments and 3D-printed discs were carried out by differential scanning calorimetry on a DSC204F1 supplied by NETZSCH-Gerätebau GmbH, Germany. All tests were performed at a heating rate of 10 °C min −1 , between −60 to 300 °C in argon atmosphere (50 mL min −1 ), using about 10 mg of sample of each composition. Heat flow data from the first heating operation was recorded.
Electron microscopy. The general homogeneity and the inner material dispersion within the PLA matrix of the produced composite filaments was examined by making use of a FEI Quanta 200F (Thermo Fisher Scientific, USA) scanning electron microscope (SEM) in high vacuum mode. The secondary and backscattered images were recorded with a 5 kV acceleration voltage. Energy Dispersive X-ray Spectroscopy (EDS) elementary maps (Fe, P, Si, C) for the "one-shot" sample were recorded at 20 kV. Selected Area Electron diffraction (SAED) patterns and transmission electron microscopy (TEM) images were obtained with a FEI Tecnai G2 electron microscope operated at 200 kV. Samples were prepared by dispersing the powder in ethanol and depositing a drop of the suspension onto a holey carbon copper grid.

Scientific RepoRtS |
(2019) 9:18031 | https://doi.org/10.1038/s41598-019-54518-y www.nature.com/scientificreports www.nature.com/scientificreports/ Mechanical characterization. As described in our previous work 42 , following solvent casting, the composite films were compression molded into 2 mm thick plates using a Doloutes hydraulic press at 160 °C during 2 minutes prior cooling down to room temperature. Afterwards, samples were cut into specimens with a punch as per ISO 527 standard for additional mechanical characterization (tensile testing). The tensile mechanical experiments were accomplished on a Llyod LR 50 K tensile machine set with a 1 KN load cell. Tests were controlled by displacement with a speed of 10 mm min −1 . Five samples of each composition (dumb bell shaped) were tested and the average values are reported.
Electrical conductivity characterization. Electrochemical impedance spectroscopy tests were done by using a MTZ-35 frequency response analyzer and an intermediate temperature system (ITS) supplied by BioLogic, France. It is worth noting that the exact same procedure than the one reported in our anterior work was respected 42 . Indeed, inductive phenomenon from the cables were avoided by performing system calibration of the empty ITS with cables. Therefore, only sample signal was measured subsequently. 200 µm thick films were punched into 5.50 mm diameter discs. The latter were introduced into a controlled environment sample holder (CESH) to perform AC impedance under air at temperatures varying from 20 °C to 50 °C (upon heating in steps of 10 °C). An excitation voltage of 0.01 V and a frequency range of 0.2 MHz to 1 Hz (20 points per decade and 10 measures per points) were applied here. Film electronic conductivities and activation energy were deduced from the Nyquist and Phase-Bode plots of the complex impedance. Conductivities were calculated from the Eq. 1: where d is the pellet thickness, A is the pellet surface area, and R is the respective resistances determined from the Nyquist and Bode plots.
Electrochemical characterization. Inside