Magnetically aligned graphite electrodes for high-rate performance Li-ion batteries

Transcription

1 ARTICLE NUMBER: Magnetically aligned graphite electrodes for high-rate performance Li-ion batteries Juliette Billaud 1+, Florian Bouville 2+, Tommaso Magrini 2, Claire Villevieille 1 *, André R. Studart 2 * + Authors have contributed equally to this work * Corresponding authors: claire.villevieille@psi.ch, andre.studart@mat.ethz.ch 1- Paul Scherrer Institut, Electrochemical Laboratory, 5232 Villigen PSI, Switzerland 2- Complex Materials, Department of Materials, ETH Zürich, 8093 Zürich, Switzerland NATURE ENERGY 1

2 Supplementary Figures Supplementary Figure 1 Comparison of the aligned and non-aligned anodes of this study (in red and blue) with literature values for the areal capacity of anodes versus C-rate. 1 2,3 References: commercial graphite anodes, Graphite SFG6 and SFG44, 3D Li4Ti5O12 4 Supplementary Figure 2 Micrograph of graphite flakes (Alfa Aesar, Graphite flake, Natural, 325 Mesh, 99.8%, metals basis) after alignment in water under a rotating magnetic field 2 NATURE ENERGY

3 Supplementary Figure 3 Setup used for casting of graphite suspensions followed by magnetic alignment of flakes Supplementary Figure 4 Schematic drawing illustrating the effect of static and rotating magnetic fields on the orientation of flakes.generated by a 400 mt Neodymium magnet. The white arrow in the bottom left corner indicates the rotation plane of the magnet. NATURE ENERGY 3

4 Supplementary Figure 5 Effect of thickness reduction by calendering on the crystallographic texture of the initially aligned sample. a. Comparison of the diffractogram obtained at different thickness reductions on an electrode with aligned graphite platelets. The blue curve corresponds to the pristine aligned electrode, whereas the red indicates the reference sample. Peaks indexed with a * correspond to the copper current collector. b. Intensity of the (002) peak in the aligned electrode after calendering divided by the intensity of the peak for a reference electrode as a function of the relative decrease in thickness imposed by calendering. 4 NATURE ENERGY

5 Supplementary Figure 6 SEM micrograph of an aligned electrode after 50 lithiation/delithiation cycles. The current collector has been removed after cycling. Supplementary Figure 7 Galvanostatic cycle for high loading (9.1 mg/cm 2 ) electrodes at C/30 rate for the two first cycles and C rate for the remainder of the test. NATURE ENERGY 5

6 Supplementary Figure 8 Zeta potential of the graphite powder in water as a function of the ph (Alfa Aesar, Graphite flake, Natural, -325 Mesh, 99.8%, metals basis). 6 NATURE ENERGY

7 Supplementary Figure 9 Structural characterisation of electrodes. a. Pictures of the aligned and reference electrodes generated after processing data obtained from the Focused Ion Beam (FIB) / Scanning Electron Microscopy (SEM) images (structures are displayed at the same scale). b. Example of diffusivity calculation indicating the boundary conditions assumed. NATURE ENERGY 7

8 Supplementary Figure 10 Streamline of the diffusive flux for two different concentration gradient directions for the reference electrode and the aligned ones 8 NATURE ENERGY

9 Supplementary Table Supplementary Table 1 Data used to calculate the tortuosity factor tensor. Aligned Not aligned Flux Direction (mol.m 2 /s) C 0 (mol/m 3 ) L 1 (m) ε D eff (m 2 /s) D 0 (m 2 /s) τ x y z x y z NATURE ENERGY 9

10 Supplementary Notes 1. Comparison of the capacity of various anodes at different charging rates We calculated the areal capacities of our electrodes and compared them in Supplementary Fig. 1 with data reported in the literature. Several important aspects are illustrated in this graph. First and most importantly, it shows that the alignment of flakes alone leads to a remarkable increase in areal capacity, even if the graphite source used is not optimised for batteries. Second, graphite flakes have been previously shown to result in high areal capacities at high rate if the particles size is optimised for lithium insertion (see SFG6 compared to SFG44 grades). This implies that flake geometries (in addition to spherical mesocarbon microbeads, MCMB-type) can also provide higher capacities that can potentially rival commercial electrodes if the formulation is further optimised with current industrial knowhow. Finally, commercial graphite anodes exhibit the highest areal capacities, which reflects the continuous incremental optimisation of these materials in industry. However, these high capacities are only possible at low rate, as the highly-loaded electrodes (i.e. with an areal capacities superior to 2 mah/cm 2 ) could not be cycled above C/3 3. The comparison between the performance of experimental anodes with battery-grade graphite flakes (SFG6 and SFG44) with commercial electrodes indicates that formulations that have been continuously optimised in industry for decades can increase the area capacity by a factor of 3 as compared to values reported for standard recipes in the open, academic literature. Such comparison further emphasises the major breakthrough achieved in our study, since a similar 3-fold increase in performance was achieved by solely aligning flakes in an orientation that facilitates mass transport of Li ions through the electrode thickness. Although further work is still needed, our approach clearly shows the great potential of controlling the architecture of the electrode alone as a simple and effective means to significantly change the areal capacity of industrially-relevant graphite anodes. 2. Effect of calendering on platelet alignment Using X-ray diffraction, we followed the alignment of the platelets after calendering an aligned electrode at various thicknesses. Changes in alignment of the flakes can be monitored accurately by measuring the intensity of the (002) peak, as this peak intensity is representative of the fraction of horizontally aligned platelets. X-ray diffraction measurements were performed at room temperature with a PANalytical Empyrean diffractometer using Cu Kα-radiation. The results are summarised in Supplementary Fig. 5. The intensity of the peak does not increase until the thickness of the electrode has been decreased by 30%, with an electrode thickness reduction of 45 µm in this case. This means that the aligned structure can be compressed and densified by 30 % without any major misalignment of the platelets. The increase in intensity observed if the thickness is reduced further indicates a progressive reorganisation of the structure, while still preserving a preferred orientation. This is clearly noticed in Supplementary Fig. 5 b if the results obtained for the aligned sample are compared with those of a reference electrode as a function of thickness reduction. 10 NATURE ENERGY

11 Those preliminary results show that in our sub-optimised case, calendering can be performed to a certain extent without strongly affecting the electrode alignment Supplementary Methods 1. Calculation of the effective diffusivity tensor from FIB-Tomography data 1.1 Mesh production and optimisation The stacks obtained from FIB-Tomography were segmented using the open source software Fiji 5 as described in the main text. The plug-in statistical region merging 6 was used with a number of merged regions of 25. Then a simple threshold was applied (Supplementary Fig. 9 a). The open source plug-in BoneJ 7 was used to extract STL files from the binary stacks with the isosurface function. The STL files represent an unscaled meshed surface, but usually with a poorly controlled vertices quality and thus need to be cleaned, scaled and optimised before use. The open source software Meshlab was employed to first remove the unconnected regions by applying the function Remove isolated pieces (wt. diameter) with a diameter of 100 pixels. Then the function Select self-intersecting faces was used to clean the mesh further. A scaling function was also applied to change the mesh size; in our case one voxel equals to 60x60x60 nm 3. The surface mesh was then imported into the open source software Gmsh 8 to produce a 3D mesh. The software native mesh optimising tools was used before exporting them as NASTRAN files Effective diffusivity calculation The procedure described here has been adapted directly from an example available in the software COMSOL Multi-physics 5.2 library. The 3D meshes were imported directly as NASTRAN files. An automatic face detection function was used to select the different entry and exit faces (Supplementary Fig. 9 b). The material properties were selected to represent lithium hexafluorophosphate (LiPF 6 ) in 1:1 EC:DEC, which is the electrolyte composition we used in our experimental measurements. A bulk diffusion coefficient DD! = 3 10!!" mm! /ss of lithium in the electrolyte was assumed. A constant concentration of CC! = 10 mmmmmm/mm! was applied on one face and an external forced convection was applied on the other side, with a mass transfer coefficient of kk = 1 mm/ss and an external concentration equal to CC!"# = 0 mmmmmm/mm!. The steady state is reached when the outward flux and concentration reaches a constant level, named respectively jj! and cc!, as shown in Supplementary Fig. 9 b. Under such condition, Fick s first law can be written as: jj! = DD!"" cc (Eq. S1) NATURE ENERGY 11

12 where jj! is the outward flux (equal to jj! = kk cc! ), DD!"" is the effective diffusivity coefficient in the considered direction and cc is the concentration gradient. Thus, DD!"" is expressed as,: DD!"" = jj!!!!!!!! = kk cc!!!!!!!! (Eq. S2) where LL! is the thickness of the mesh in the considered direction. Finally, DD!"" can now be expressed as a function of the bulk diffusion coefficient DD! : DD!"" =!! DD! (Eq. S3) with εε being the electrode porosity calculated from the mesh volume in COMSOL, and ττ being the tortuosity factor, which in this case is an adjustable parameter representing the electrode morphology in the considered direction 9. The data resulting from the analysis of the two stacks shown in Supplementary Fig. 9 are summarised in Supplementary Table 1 and plotted in Fig. 3 of the main manuscript Streamline of the diffusive flux To further illustrate the modification of mass transport in our structure, we used the streamline option of the Comsol Software to plot a number of lines representing the mass flux direction for two cases, one where the concentration gradient was along the x direction, and one where it was along the z direction, for both electrodes (cf. Supplementary Fig. 10). Indeed, the lines are straight and parallel to each other when the tortuosity factor is low, and randomly oriented when the tortuosity factor is high. 12 NATURE ENERGY

13 Supplementary References 1. Gallagher, K. G. et al. Optimizing Areal Capacities through Understanding the Limitations of Lithium-Ion Electrodes. J. Electrochem. Soc. 163, A138 A149 (2016). 2. Buqa, H., Goers, D., Holzapfel, M., Spahr, M. E. & Novák, P. High Rate Capability of Graphite Negative Electrodes for Lithium-Ion Batteries. J. Electrochem. Soc. 152, A474 (2005). 3. Heß, M. & Novák, P. Shrinking annuli mechanism and stage-dependent rate capability of thin-layer graphite electrodes for lithium-ion batteries. Electrochim. Acta 106, (2013). 4. Sun, K. et al. 3D printing of interdigitated Li-ion microbattery architectures. Adv. Mater. 25, (2013). 5. Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, (2012). 6. Nock, R. & Nielsen, F. Statistical region merging. IEEE Trans. Pattern Anal. Mach. Intell. 26, (2004). 7. Doube, M. et al. BoneJ: Free and extensible bone image analysis in ImageJ. Bone 47, (2010). 8. Geuzaine, C. & Remacle, J.-F. Gmsh: A 3-D finite element mesh generator with builtin pre- and post-processing facilities. Int. J. Numer. Methods Eng. 79, (2009). 9. Doyle, M., Newman, J., Gozdz, A. S., Schmutz, C. N. & Tarascon, J.-M. Comparison of Modeling Predictions with Experimental Data from Plastic Lithium Ion Cells. J. Electrochem. Soc. 143, 1890 (1996). NATURE ENERGY 13