3D Printed Rechargeable Aqueous and Non-Aqueous Lithium-Ion Batteries: Evolution of Design and Performance

Three versions of coin type full cells with architected porous current collectors were designed using SideFX Houdini Apprentice software and exported as STL files. The version I (V1) cell was equipped with a 3D printed gasket, version II (V2) was gasketless with a monolithic current collector/case, and version III (V3) had separate current collector pieces attached to the case after metal coating (Fig. 1). The cells were modeled with a modified diamond cubic (DC) or body-centered cubic (BCC) microlattice current collector geometry (Fig. 2A). These ordered porous structures, which function as the current collectors, have two benefits: 1) they allow defined total volume recesses for active material slurry without contributing to the thickening of the cell. 2) each "pocket" of active material is surrounded by current collector ensuring good and equal electrical connectivity through each electrode to minimize any electrochemically inactive "dead weight" active slurry material for a given gravimetric/volumetric mass loading.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The ordered porous current collector lattice was rotated in software as appropriate, and the topmost edges were flattened in software prior to printing to ensure good even contact between cathode and anode sides and the electrolyte-soaked separator. The STL models were processed using Formlabs PreForm slicer software—scaled, duplicated, and oriented in accordance with the printer instruction manual. Supports were automatically generated by software with manual editing where necessary. All cell parts were printed using Formlabs Form 2 desktop SLA 3D printer with 25 μm layer thickness.

Clear V4 and High Temp V2 methacrylate-based resins (Formlabs) were selected for printing due to their relatively good solvent resistance, High Temp V2 being also heat resistant, with full details of the resins shown in Table I. Upon completion of the printing, the objects attached to the build platform were rinsed at room temperature with isopropyl alcohol (IPA) in the Form Wash automated wash station to remove uncured resin. After washing, the parts were removed from the build platform, dried with an air blower, and post-cured in the Form Cure station (λ = 405 nm illumination, Formlabs) with the supports attached. Wash and post-cure parameters used are shown in Table II. Supports were carefully detached from the post-cured objects; the latter were polished using 800 and 1200 grit sandpaper to remove support touchpoint dents. After rinsing in IPA and drying, the current collector containing parts were fully coated with gold or nickel (nominal layer thickness 150–200 nm) using a Quorum Q150T S sputter coater equipped with a Ted Pella 57 × 0.1 mm Au or Ni target.

Iron phosphate dihydrate (FePO₄·2H₂O), lithium manganese oxide (LiMn₂O₄, spinel, <0.5 μm particle size, >99%), N-Methylpyrrolidone (NMP, anhydrous, 99.5%), Poly(vinylidene fluoride) (PVDF), and lithium bis(trifluoromethanesulfonyl)imide (LiTFSI, anhydrous, 99.99% trace metals basis) were purchased from Sigma–Aldrich and used as received. Lithium cobalt oxide (LiCoO₂), graphite, Super-P carbon black and carboxymethyl cellulose—styrene-butadiene rubber binder (CMC-SBR, M.W. 400,000 g mol⁻¹) were purchased from MTI Corporation. Formlabs Clear Resin V4 (FLGPCL04) and High Temp resin V2 (FLHTAM02) were used for vat polymerization 3D printing. A 1.0 mol L^–1 lithium hexafluorophosphate solution in ethylene carbonate/diethyl carbonate (50/50 v/v, Sigma–Aldrich) was used as electrolyte in non-aqueous batteries. Gorilla Epoxy two-component adhesive, Henkel Loctite hot melt glue gun sticks, and Prevest DenPro Fusion Flo light cured dental nano hybrid cement were tested as sealing materials for the cells. Whatman GF/D glass microfiber filters and Celgard 2325 PP/PE/PP trilayer membrane were used as separators in aqueous and non-aqueous batteries, respectively.

Anode and cathode material slurries for aqueous batteries were prepared by ball milling super-P carbon black with FePO₄·2H₂O (FPO) and LiMn₂O₄ (LMO), respectively, using a high-speed 3D ball mill (MTI corporation) with a maximum speed of 1200 rpm. Later, each powder was mixed with CMC–SBR and water suspension using a compact vacuum mixer (MTI corporation). The final slurry consists of 80% w/w active material (FPO or LMO), 10% Super P carbon black and 10% CMC–SBR binder with solid content ranging from 5%–10%. Ethanol was used to adjust the surface tension of the slurry mixture in order to make it spreadable inside the porous 3D substrates. Electrodes were prepared by drop casting each slurry into 3D printed Au-coated substrates and dried inside oven at 65 °C for 12 h. A gasket was attached where applicable (V1 cell), and a glass fibre separator (Whatman GF/D) was placed between cathode and anode parts (Fig. 1), 10 M LiTFSI aqueous electrolyte was added, and the resulting cell was sealed around the edges with one of the three tested sealants. The cross-section area of each 3D printed electrode was 3.14 cm² while the loading ratio between FPO-based anode and LMO-based cathode was 2.0. When the dental cement was used as a sealant, the initial curing was performed inside a glovebox (to avoid oxygen exposure) using a handheld LED UV torch (λ = 395 nm), then the assembled cells were additionally cured in the Form Cure curing device at room temperature for 5 min.

Active material slurries for non-aqueous cells were prepared similarly to the above procedure with a number of differences related to the non-aqueous nature of the system. Graphite and lithium cobalt oxide were used as an anode and cathode material, respectively; N-methylpyrrolidone and PVDF were used as a solvent and binder for the slurry preparation. The cathode material slurry composition (LCO:super P:binder) was 80:10:10 w/w/w. Anode material slurry contained graphite, binder (90:10 w/w) and solvent. Celgard 2325 membrane separator was cut in 16 mm diameter circles and placed between anode and cathode in non-aqueous batteries. Electrolyte addition (1 mol l⁻¹ LiPF₆ in EC/DEC 1:1 v/v) and the final assembly of non-aqueous cells was performed inside a dry glovebox under Ar atmosphere (<0.1 ppm O₂/H₂O).

The quality assessment of 3D geometric shapes was carried out using scanning electron microscope (SEM) images obtained with an FEI Quanta 650 SEM at spot size 3 and 20 kV beam voltage. Additionally, Energy Dispersive X-ray (EDX) analysis was employed to study the morphology and distribution of gold coating and the distribution of active material slurry within the pores of the current collector. Electrical conductivity evaluations were performed using a two-probe technique, and I − V data was gathered with a Keithley 2612B source meter at an integration time of 50 ms per point between 0 and 0.2 V. A consistent distance of 20 mm was maintained between the probes for all tests. Readings were taken across the porous side, the flat side, and between opposing planes of each electrode. Raman scattering analysis was performed on both uncoated and metal-coated electrode surfaces using a Renishaw InVia Raman spectrometer, equipped with a 30 mW Ar⁺ laser at an excitation wavelength of 514 nm. A 40x lens was used to focus the laser, and a RenCam CCD camera collected the data.

Battery cycling and electrochemical measurements were performed using BioLogic VMP3 multichannel potentiostat controlled by EC-Lab software. Aqueous 3D printed cells in this study were cycled in the potential range of 0–1.3, 1.35, 1.4 V vs Li/Li⁺ at various current densities in CC/CC or CCCV/CC modes (CC = constant current; CV = constant voltage). During the first formation cycle, non-aqueous batteries were charged to 4.0 V at 0.05 C constant current; the cycling was subsequently performed in the voltage range 3.0–4.2 V at various current densities in CCCV/CC mode.

The suitability of nickel and gold as coating materials for 3D printed conductive substrates was tested. First, electrical conductivity measurements were performed on Au-coated and Ni-coated version II substrates, which showed that Au-coated substrates had significantly lower resistance (1.6–2.9 Ω) compared to Ni-coated substrates (29–96 Ω). Both clear and high temp resin substrates demonstrated very close conductivity results for all tested metal coating and lattice type combinations. For Au-coated substrates there was no noticeable difference observed between body centered cubic (BCC) and diamond cubic (DC) lattice. However, for Ni-coated substrates, DC lattice showed higher resistance than BCC lattice whenever one of the terminals was connected to the porous substrate (Table III).

The results suggest that Au-coated substrates provided better electrical conductivity compared to Ni-coated current collectors. The Au-coated substrates were also found to be significantly more stable after immersion in 10 M LiTFSI based aqueous electrolyte for 24 h. Raman scattering and imaging measurements showed negligible change to the resistance values of Au-coated substrates before and after electrolyte treatment, whereas Ni-coated substrates practically became significantly more resistive and in regions where delamination of Ni occurred, were non-conductive. Finally, it was observed that there was no significant difference in resistance for Au-coated substrates printed from two different polymer resins.

High resolution SEM images along with EDX analysis were used to examine the nickel and gold coating on the surface of polymer substrates (Fig. 2B). It was observed that both nickel and gold completely covered the surface of the substrates and on the surfaces within each of the pores to ensure electrical connection with the infilled slurry material. For all tests in each version of the cell reported here, Au-coated porous current collectors with DC porous internal geometry were used.

EDX analysis of the current collector showed the uniform metallization by Au, shown in Fig. 2C. Comparison of scanning electron microscopy (SEM) images of the cross-sectional view of gold-coated substrate before and after the introduction of cathode material reveal a tightly packed arrangement of active material throughout the lower layers of the porous 3D electrode, and relatively uniform overcoat of the slurry across the infilled porous current collector, as shown in Fig. 2D.

In aqueous systems, both printing resins (clear and high temp) exhibited equally good electrolyte resistance with the testing results independent of the resin type. However, in the case of non-aqueous batteries, only the high temp resin was stable enough to maintain the shape and integrity of electrodes throughout the cell preparation procedure. The reason is that cathode material slurry contained N-methyl-2-pyrrolidone, which caused swelling of the clear resin polymer (summarized in Table II), but did not affect the high temp resin. Additionally, the electrode preparation in the case of non-aqueous systems involved solvent evaporation at higher temperatures than in the case of aqueous cells. These reasons dictated the choice of resins for fabrication of aqueous and non-aqueous batteries: the former were predominantly printed with clear resin, the latter were assembled using high temp prints only.

Version I design was inspired by a conventional coin cell structure with an insulating gasket placed between cathode and anode current collectors (Fig. 1A). Photos of 3D printed Version I cells parts are shown in Fig. 3. After preliminary tests with the aqueous Li-ion system it was found that although the gasket did perform as a good insulator, it also introduced three major challenges. The first issue was the brittleness of both types of cured resins used for printing. The printed gasket walls were relatively thin and easily broke upon such handling operations as detaching from supports and assembly of the cells. The cell case parts also occasionally cracked during cell assembly when some additional pressure was applied during assembly of the entire battery and all its parts.

Zoom In Zoom Out Reset image size

The second issue with models that use separate gaskets was the limitation of Vat-P 3D printing resolution, which puts a lower limit on the thickness of some case elements and the gasket. This leads to an larger overall size of the assembled cell and a large unutilized overall volume. Finally, even Vat-P 3D printing, which is considerably better in render quality and resolution compared to fused deposition modelling (FDM for example), has limitations due to imperfections and small size variations in the resulting prints. When practitioners use this printing method, care should be taken to inspect the prints to avoid large and uneven electrode gaps, poor sealing, and difficulties in the gasket/case fitting. We managed to assemble and test several Version I batteries using aqueous chemistry. The electrochemical testing results are presented in Figs. 4 and 5, to be discussed later. That said, the aforementioned issues prevented any further testing of the V1 cells, and some modifications were applied in the Version II design.

Zoom In Zoom Out Reset image size

The Version II printed coin cell was designed to be assembled without a separate printed gasket. In doing so, several printing and assembly issues were remedied including all the gasket supporting elements. A separator was used as an electronic insulator instead (Fig. 1A). By implementing this change, we were able to significantly decrease the cell thickness and overall coin cell volume and weight, leaving the electrode area almost unchanged. Reducing the number of elements considerably simplified the handling and assembling procedures of each cell.

As discussed earlier, a sputtered nickel coating proved to be readily oxidized in air, which reduced the surface conductivity by an order of magnitude. For this reason, the majority of electrochemical cycling tests were conducted with gold-coated current collectors only, which are stable both in air and at the cycling conditions in both electrolytes in their respective voltage ranges. Several Version II aqueous lithium-ion batteries were assembled and cycled. The electrochemical results are discussed in below. Despite the successful testing Version II with aqueous electrolytes, all attempts to achieve even a single charge/discharge cycle in non-aqueous electrolytes failed. Post-mortem analysis showed that the main reason for the failure was the exfoliation of the gold coating at the edges of the case/current collector and contact areas under the influence of a relatively low-viscosity organic electrolyte solution. There was an apparent exfoliation and thinning of the gold layer around the cell's outer regions after the addition of electrolyte. To implement a Vat-P printed cell that was stable in both aqueous and organic electrolytes, the Version III cell design was conceived. In this version, current collectors are separated from the rest of the case, and a circular wall is added to one of the case parts in order to improve sealing (cf Fig. 1A). The electrode surface area is slightly reduced in Version III cells. This approach allows us to reduce the amount of metal used for coating, avoid any contact between the gold coating during the cell assembly, and improve cell sealing. With the Version III layout we were able to successfully cycle the cells in both aqueous and non-aqueous batteries, as we be discussed in the next section.

In developing resin or plastic-based rechargeable aqueous and non-aqueous Li-ion batteries, new issues emerge when the plastic needs to be engineered or modified to maintain, as close as possible, some useful attributes of metal casings. While polymeric casings are much lighter and benefit overall cell-level energy density, polymer 3D printed cases cannot usually withstand high pressure and can crack during assembly or charging. This was the case with Version I and Version II designs during our testing. In the absence of a conductive resin that at the time of writing is not available, the design is limited by a conductive surface coating. Coating integrity is a function of geometric complexity in printed design and can potentially create lower electrically conductive pathways at the high-curvature regions due to erosion by electrolyte and lower accessibility during vapor or electroless deposition. However, modifications outlined for Version III were successful overall for both aqueous and non-aqueous rechargeable cells.

Since 3D printed cells cannot be crimped or compressed like coin cells during assembly, alternative sealing methods need to be developed to maintain structural integrity, avoid electrolyte leakage and evaporation, and to prevent outside atmosphere ingress particularly for organic electrolyte containing cells. We did extensive testing of three different sealants: (1) epoxy resin, (2) EVA hot-melt adhesive, and (3) uv-cured dental hybrid cement. The best performance was achieved with dental cement due to its extremely high strength and solvent resistance, as well as excellent adhesive properties. In order to avoid the formation of oxygen-inhibited layer, the sealing is best performed in an oxygen-free atmosphere. The option to use the same uncured resin as a sealant was ruled out because of insufficient viscosity of the resin.

Inexpensive EVA adhesives have a low creep resistance ⁴⁰ and a poor adhesion to the photopolymerized PMMA-based resins used for printing, which adversely impacted the sealing quality and limited their application in our experiments. Hot-melt adhesive requires a significant time to cool and solidify after application. This can make it challenging to achieve consistent and reliable seals. Additionally, EVA hot-melt glue has poor solvent and chemical resistance to organic electrolytes and active compounds used in lithium-ion batteries. Plasticizers used in the glue compositions may contaminate the electrolyte and cause battery performance degradation.

While the epoxy glue adhesive demonstrated reasonably good sealing quality in aqueous systems and strong adhesion to the cell surface, it had some drawbacks in non-aqueous batteries such as limited resistance to organic solvents, relatively long curing time, and sensitivity to component mixing ratio. Also, the components of epoxy resin can contaminate organic electrolytes, deteriorate the battery performance, and ultimately lead to its failure. Epoxy glues can become brittle over time, which may lead to seal failure and potential leaks. The ceramic-based dental resin, once cured, was a very effective and consistent sealant for 3D printed coin cells for aqueous and non-aqueous internal chemistry.

All of the designs of Vat-P printed coin cell batteries discussed above were electrochemically tested using at least one kind of Li-ion battery chemistry (aqueous or non-aqueous). Due to the limitations of some materials discussed above, the first two versions (Version I and Version II) were cycled as aqueous cells only.

The electrochemical behavior and performance of 3D printed Li-ion batteries was tested using galvanostatic charge/discharge cycling in constant current/constant voltage (CCCV) or constant current (CC) charging mode. Differential capacity plots obtained from the galvanostatic charge-discharge curves for the 3D printed Li-ion battery showed two prominent pairs of oxidation and reduction peaks (Fig. 4A). These peaks represent the lithiation and delithiation of the anode FePO₄ and cathode LiMn₂O₄, respectively, and can be attributed to two reversible electrochemical reactions: 2Li_0.5Mn₂O₄ + Li⁺ + e^– ⇆ 2LiMn₂O₄ and 4MnO₂ ⇆ 2Li_0.5Mn₂O₄ + Li⁺ + e^–. ⁴¹

Discharging was performed at constant current in all cases. Self-discharge of the charged cells was evaluated by measuring the variation of the open circuit voltage (OCV) for 12 h. This measurement is important for 3D printed cells and cells made from non-traditional current collector geometries and materials to quantify the stability of the cell at a high state of charge. The results are summarized in Fig. 4B. As can be seen, in the case of CC charging, a rapid and deep decline in the cell voltage is observed, which indicates that chemical and diffusion processes tend to equilibrium for some time after the charging ends. This can be due to the increased thickness of active material layers inside the cell, hence the kinetic delay in reaching equilibrium. Nevertheless, this quite significant drop in the cell voltage (>20%) translates to less than 5% drop in capacity of the battery in the case of the Version I cell charged to 1.2 V. Utilizing CCCV charging mode (0.8–1.35 V; 0.8 V—1.5 V) significantly improved self-discharge behavior, where both Version I and Version II cells maintained 86% and 92% of OCV after 12 h. This charging method allows better utilization of active material and accelerating transition to the equilibrium state.

Figure 4C shows the first 3 cycles charge/discharge profiles for different versions of aqueous coin cell. The profile shape is in good agreement with the literature data ¹⁴ for this aqueous system in standard-type cells, as a function of the capacity of the entire cell, not just the specific capacity of one of the electrode active materials. The cells are very stable, and we note an improvement overall in a larger voltage window up to 1.35 V. Specific discharge capacity and coulombic efficiency plots for all cell versions with the aqueous electrolyte in different voltage cutoff ranges are summarized in Figs. 5A, 5B. In general, the specific capacity of the cell briefly increases during first several cycles reaching peak values of 12, 30, and 27 mAh g⁻¹ for Versions I, II and III cells, respectively. However, a steady decay in cell capacity is observed after 30th–50th cycle. Low coulombic efficiencieds obtained in the first several cycles of the 3D printed battery could indicate the intensive formation of SEI layers inside its highly porous 3D architecture, since none of the material were prelithitated or electrochemical pre-conditions prior to use at the stable cycling voltage. After several initial formation cycles, coulombic efficiency tends to 95%–100%, which indicates the reversible nature of electrochemical processes during charge/discharge cycling at this stage. Gradual increase in capacity can be explained by the fact that thick and porous 3D electrode allows complete utilization of active material, but there are kinetic limitations due to the thickness of the electrodes and the volume of material in each of the current collector pores. As the cycling proceeds, more material is available for lithiation and delithiation, thus increasing the capacity. Once the capacity of the cell achieved the maximum value, it remains stable for a prolonged cycling period before gradual decline. We believe that the performance levels of the 3D printed Li-ion batteries with thick, porous electrodes we show here do not fully showcase their true capability and are an unoptimized lower bound. Additional research and advancements are necessary to fully harness the potential of the active materials in these types of cells.

Zoom In Zoom Out Reset image size

The design ensures that the conductivity of 3D printed porous electrodes is improved and behaves as a metallic coin cell casing in this regard. This was also confirmed using rate capability testing (Fig. 5C). The shape of the plots was almost independent on the battery version and was defined by the active material composition and current collector geometry (which was the same in all cell versions). We suggest that the good rate capability of 3D printed lithium-ion aqueous batteries is the result of highly interconnected network that provides multiple pathways for electronic conductivity within the porous current collector.

Table IV presents a comparison of the performance characteristics for both conventional and 3D printed aqueous lithium-ion cells featuring FPO/LMO active materials. When considering gravimetric capacities (based on the total mass of the cell), the 3D printed cell surpasses the traditional coin cell. This is because the 3D printed lithium-ion cells are approximately half the weight and deliver higher capacities at elevated rates. However, when dense and thick electrodes are used in the 3D printed aqueous, it can restrict the full use of the active material, possibly due to inadequate wetting of the 3D printed electrode.

Internal resistance measurements once again confirmed the advantage of the Version III design compared to Version I and Version II. The average internal resistance for aqueous batteries was 3.2 ± 0.8 Ω (Version I), 3.2 ± 0.8 Ω (Version II) and 0.4 ± 0.3 Ω (Version III). As mentioned above, non-aqueous 3D printed cells using carbonate-based electrolytes were successfully assembled and tested using the Version III cell design only. The electrochemical response is presented in Fig. 6. The results show that Version III 3D printed coin cells reached the specific capacity of ∼115 mAh g⁻¹ caculated using the mass of the LCO active materials, which is only ∼20% less than the maximum reversible capacity of LoCoO₂. ⁴² The entire cell itself delivered ∼1.14 mAh g⁻¹.

Zoom In Zoom Out Reset image size

The capacity per unit footprint area was higher for the 3D printed cells than for conventional coin cells due to the increased surface area and three-dimensional structure of the porous current collector. Similar to the aqueous batteries, the gravimetric capacity per unit mass of the assembled cell (∼12 mAh cm⁻²) was higher than for conventional coin cells (∼8–10 mAh cm⁻²) due to the low case material density. Rate capability measurement results showing the capacity relative to the capacity at a rate of 1 C were on par with conventional LCO coin cells (Fig. 6D). Although the maximum achieved specific capacity was somewhat lower than the theoretical capacity of LCO material, we believe that the performance of Vat-P or SLA 3D printed batteries can be significantly improved, and it already compares well compared to coin cells using LCO with graphite anode material. The response reported here constitutes the state of the art for non-aqueous Vat-P printed cells using these electrode materials.