Cell-Internal Contacting of Prismatic Lithium-Ion Batteries Using Micro-Friction Stir Spot Welding

: The reliable production of high-quality lithium-ion battery components still poses a chal-lenge, which must be met to cope with their rising demand. One key step in the production sequence is the process of cell-internal contacting, during which the electrode carrier foils of the anode and the cathode are joined with the arrester. This is usually done with ultrasonic or laser beam welding. Both joining processes, however, show limitations concerning the quality of the weld. This paper presents a new approach for cell-internal contacting by using micro-friction stir spot welding. Welding experiments were conducted in which joints with high mechanical strengths were produced. It was also shown that large stacks with foil numbers of 100 can be joined in only a few tenths of a second. The process is therefore especially of interest for the fast production of large-scale battery cells or other new types of high-energy-dense battery cells.


Introduction and State of the Art
The increasing demand for lithium-ion batteries [1] (i.e., for electric vehicles) requires fast and high-quality production processes for its components. Lithium-ion battery packs consist of single battery cells [2] for which the production chain encompasses a large number of steps [3] (pp. 212-219). One of these is the cell-internal contacting (example in Figure 1). During this step, the uncoated parts of the stacked or rolled collector foils are joined with the arrester [2] (also called collector [4,5]). The quality of the joint is crucial, since it is part of the electrical circuit of the cell [6] and because rejects late in the production chain increase costs [7].
Currently, ultrasonic welding (USW) and laser beam welding (LBW) are industrially applied for cell-internal contacting [3] (p. 217). For both processes, however, quality reducing defects, such as ruptured foils (USW and LBW) [8,9] pores (only LBW) [5,8], and a low reproducibility (USW [10]), have been reported. Another problematic issue is the possible damaging of the cell due to process induced vibrations (USW [2]) or spatter and fumes (LBW [5]). With regard to large-scale battery cells with a high layer number [11] and therefore high material costs [7,12], the process stability is of high importance. This is especially important with the cell-internal contacting; approx. 80% of the total manufacturing costs of battery cells are accumulated [12]. Not only a fast, but also a robust process is needed for large-scale cells. Robust in this case means that welds with joint properties, which meet the defined requirements, can be generated even when disturbances occur (e.g., imprecise fit-up of the foils). With USW and LBW, however, the welding time or defects and spatter (LBW [5]) increase with the number of welded foils, since higher joining energies are necessary for welding larger parts [13]. vp) into the joining partners. The tool dwells there for a specified time td to increase the material softening and flow within the weld (second step). Afterwards (third step), the tool is retracted (retract speed vr), thereby leaving a characteristic negative imprint and flash on the surface. The process is highly automated and is typically monitored by using inline measured signals of the axial force Fz and the spindle torque Mz (process variables) [18][19][20].  [11] and (b) the process sequence for cell-internal contacting using μFSSW in multi-sheet butt joint configuration. The arrows indicate the direction of movement.
A recent study [21] revealed the high potential of the process for cell-internal contacting. Gera et al. [21] joined stacks consisting of 50 commercially pure aluminium foils (no standard given) with one arrester tab (aluminium alloy 2024-T3, no standard given) on each side of the stack in a multi-sheet lap joint configuration. They employed a special three-piece welding tool, which included an additional circular clamping ring outside the rotating shoulder ("refill FSSW" [14]) and with which defect free spot welds with good mechanical properties were produced.
The results show that it is feasible to weld foil stacks with arrester tabs using μFSSW. The welding experiments were, however, not conducted with commonly used cathode and anode materials (pure aluminium and copper), which impedes an assessment of the process for cell-internal contacting, and results in a need for further investigations.

Objective of the Investigations
The objective of the investigations reported in this paper was a holistic assessment of the μFSSW process for cell-internal contacting of prismatic lithium-ion battery cells. For this, welding experiments were conducted, and the results discussed with regard to the following requirements (R1-R8) taken from literature: R1. high process stability and no welding defects [5]; R2. no spatters or particles to prevent short circuits inside the cell [22,23]; This analysis shows that LBW and USW show disadvantages for the application of cell-internal contacting, especially for large-scale cells. One possible alternative joining technology is micro-friction stir spot welding (µFSSW). The process is derived from friction stir spot welding (FSSW), which is used in the aerospace, automotive, and railway industry [14]. It is known for the high quality of the welds [15] and the high process robustness [16]. The prefix "micro" (µ) is added when workpieces with dimensions below 1 mm (e.g., foils) are joined [17].
µFSSW is a solid-state process (like USW) with which the joining partners are welded through frictional heat and material flow, and not through melting (like LBW). In Figure 1b, the process sequence is shown for a multi-sheet butt joint configuration. In the first process step, a rotating tool consisting of a shoulder and a probe is plunged (plunge speed v p ) into the joining partners. The tool dwells there for a specified time t d to increase the material softening and flow within the weld (second step). Afterwards (third step), the tool is retracted (retract speed v r ), thereby leaving a characteristic negative imprint and flash on the surface. The process is highly automated and is typically monitored by using inline measured signals of the axial force F z and the spindle torque M z (process variables) [18][19][20].
A recent study [21] revealed the high potential of the process for cell-internal contacting. Gera et al. [21] joined stacks consisting of 50 commercially pure aluminium foils (no standard given) with one arrester tab (aluminium alloy 2024-T3, no standard given) on each side of the stack in a multi-sheet lap joint configuration. They employed a special three-piece welding tool, which included an additional circular clamping ring outside the rotating shoulder ("refill FSSW" [14]) and with which defect free spot welds with good mechanical properties were produced.
The results show that it is feasible to weld foil stacks with arrester tabs using µFSSW. The welding experiments were, however, not conducted with commonly used cathode and anode materials (pure aluminium and copper), which impedes an assessment of the process for cell-internal contacting, and results in a need for further investigations.

Objective of the Investigations
The objective of the investigations reported in this paper was a holistic assessment of the µFSSW process for cell-internal contacting of prismatic lithium-ion battery cells. For this, welding experiments were conducted, and the results discussed with regard to the following requirements (R1-R8) taken from literature: R1. high process stability and no welding defects [5]; R2. no spatters or particles to prevent short circuits inside the cell [22,23]; R3. high repeatability of the weld quality [22]; R4. bonding of all foils and no damaged foils [22]; R5. no thermal damaging of temperature sensitive cell components [22]; R6. low electrical resistance of the joint [22];

Materials and Methods for the Experimental Investigations
The welding experiments were conducted with stacks of 30 electrode carrier foils and two thin arrester bars. A sketch of the workpiece configuration is shown in Figure 2 and the materials and dimensions are given in Table 1. For each workpiece, the foils were cut to size, stacked, and welded with the arrester bars in a multi-sheet butt joint configuration. This configuration was derived from a design with two thin arrester bars for each electrode, as shown by Lundgren et al. [24] and Kleiner et al. [4].
The welding experiments were conducted with stacks of 30 electrode carrier foils and two thin arrester bars. A sketch of the workpiece configuration is shown in Figure 2 and the materials and dimensions are given in Table 1. For each workpiece, the foils were cut to size, stacked, and welded with the arrester bars in a multi-sheet butt joint configuration. This configuration was derived from a design with two thin arrester bars for each electrode, as shown by Lundgren et al. [24] and Kleiner et al. [4].  Like LBW [9], μFSSW needs sufficient clamping to restrain the delicate foils and arrester bars from deforming during the process. The design of the clamping system is, together with the applied fit-up process, described in Appendix A.1 and shown in Figure  A1. Two different tools (tool 1 without a probe and tool 2 with a probe) were designed (see Appendix A.2, Figure A2) with which the welding experiments were conducted. For each tool (1 and 2) and material (aluminium and copper), four spot welds were produced on one workpiece, resulting in four spots produced by tool 1 and four spots produced by tool 2. This was done in order to analyze the reproducibility of the process (R3). The welding machine setup and the applied process parameters are also described in Appendixes A.1 and A.2.   Like LBW [9], µFSSW needs sufficient clamping to restrain the delicate foils and arrester bars from deforming during the process. The design of the clamping system is, together with the applied fit-up process, described in Appendix A.1 and shown in Figure A1. Two different tools (tool 1 without a probe and tool 2 with a probe) were designed (see Appendix A.2, Figure A2) with which the welding experiments were conducted. For each tool (1 and 2) and material (aluminium and copper), four spot welds were produced on one workpiece, resulting in four spots produced by tool 1 and four spots produced by tool 2. This was done in order to analyze the reproducibility of the process (R3). The welding machine setup and the applied process parameters are also described in Appendices A.1 and A.2.

Methods for Evaluation
Evaluation criteria were derived for each requirement (R1-R8) and are listed in Table 2. The testing setups are described in Appendix B. Established evaluation methods, such as visual inspection and mechanical testing, were applied and adapted to the sample configuration.

Results and Discussion
The welding experiments were conducted, and the produced samples were analyzed as described in Section 2 and Appendices A and B. A summary of all collected data is given in Table A1 in Appendix C. The results from the acquired data are presented and discussed in the following section.

Visual Inspection and Weld Topographies (R1, R2 and R4)
First, a visual inspection (see Appendix B.1) of all welds was conducted. Sound welds ( Figure 3) without defects (R1) were produced for aluminium and copper with both tools. All foils were connected and remained unruptured (R4). As expected, the tools left circular imprints on the surfaces of the parts. With tool 2, deep indentations were caused by the probe in the center of the spots. The weld surfaces for copper were generally more uniform than for aluminium. Characteristic surface galling (tool 1) and high flash (tool 2, max. flash heightsĥ f between 0.37 mm and 2.03 mm) were noticed for the aluminium spot welds. The visual inspection of tool 2 revealed the cause to be adhered aluminium material at and around the base of the probe. The material adhesion and particle detachment might be reduced by coating the tool's surface (e.g., with TiB 2 [30]).
No particle detachment from the workpieces was otherwise noticed during the welding experiments (R2). Aside from this, all other welds featured max. flash heightsĥ f below 0.39 mm, which is less than the height of the indentations in USW (approx. 0.7 mm) as reported by Shin et al. [29].
It is to be summarized that defect free welds (R1) can be produced without particle detachment (R2) by using appropriate tools (tool 1 for aluminium and both tools for copper). All welding samples showed completely attached and unruptured foils (R4).  No particle detachment from the workpieces was otherwise noticed during the welding experiments (R2). Aside from this, all other welds featured max. flash heights ĥf below 0.39 mm, which is less than the height of the indentations in USW (approx. 0.7 mm) as reported by Shin et al. [29].
It is to be summarized that defect free welds (R1) can be produced without particle detachment (R2) by using appropriate tools (tool 1 for aluminium and both tools for copper). All welding samples showed completely attached and unruptured foils (R4).

Repeatability of the Process (R3)
The topographical data (see Appendix B.2) of the four spot welds were also used to assess the repeatability of the process (R3, Figure 4a). Using tool 1 for aluminium, low standard deviations (SD) of the flash heights ĥf were achieved (0.09 mm) with a coefficient of variation (CV) of 30%. A trend in the data was noticed with tool 2 for aluminium, which is due to the described material adhesion on this tool (see Section 3.1). With copper, the lowest SD (0.01 mm) and CV (17%) were produced by using tool 2.

Repeatability of the Process (R3)
The topographical data (see Appendix B.2) of the four spot welds were also used to assess the repeatability of the process (R3, Figure 4a). Using tool 1 for aluminium, low standard deviations (SD) of the flash heightsĥ f were achieved (0.09 mm) with a coefficient of variation (CV) of 30%. A trend in the data was noticed with tool 2 for aluminium, which is due to the described material adhesion on this tool (see Section 3.1). With copper, the lowest SD (0.01 mm) and CV (17%) were produced by using tool 2. No particle detachment from the workpieces was otherwise noticed during the welding experiments (R2). Aside from this, all other welds featured max. flash heights ĥf below 0.39 mm, which is less than the height of the indentations in USW (approx. 0.7 mm) as reported by Shin et al. [29].
It is to be summarized that defect free welds (R1) can be produced without particle detachment (R2) by using appropriate tools (tool 1 for aluminium and both tools for copper). All welding samples showed completely attached and unruptured foils (R4).

Repeatability of the Process (R3)
The topographical data (see Appendix B.2) of the four spot welds were also used to assess the repeatability of the process (R3, Figure 4a). Using tool 1 for aluminium, low standard deviations (SD) of the flash heights ĥf were achieved (0.09 mm) with a coefficient of variation (CV) of 30%. A trend in the data was noticed with tool 2 for aluminium, which is due to the described material adhesion on this tool (see Section 3.1). With copper, the lowest SD (0.01 mm) and CV (17%) were produced by using tool 2. In addition to the topographical data, the process variable data (see Appendix B.3) was analyzed (R3). The measurement data of the max. axial forces Fz and the max. spindle torques M z are shown in Figure 4b,c. For aluminium, the lowest values for the AM (314.14 N), the SD (34.09 N) and the CV (11%) of the max. axial forces Fz were achieved with tool 1. As with the topographical data, a trend in the data was noticed for tool 2 (only aluminium). The max. flash heights ĥf correlated significantly (p-value below 0.05 [27], p. 74) with aluminum copper In addition to the topographical data, the process variable data (see Appendix B.3) was analyzed (R3). The measurement data of the max. axial forcesF z and the max. spindle torquesM z are shown in Figure 4b,c. For aluminium, the lowest values for the AM (314.14 N), the SD (34.09 N) and the CV (11%) of the max. axial forcesF z were achieved with tool 1. As with the topographical data, a trend in the data was noticed for tool 2 (only aluminium). The max. flash heightsĥ f correlated significantly (p-value below 0.05 [27], p. 74) with the max. axial forcesF z with a Pearson correlation coefficient of 0.9980. This indicates that the axial forceF z might be suitable for inline control of the weld surfaces.
For copper, the AMs of the max. axial forceF z and the max. spindle torquesM z were generally higher than for aluminium, which is due to its higher material strength [31]. Using tool 1, the lowest AM of the max. axial forceF z (1481.19 N) and using tool 2, the lowest SD (28.06 N) and CV (1%) were reached for copper.
Based on this analysis, it can be summarized that a high reproducibility (R3) can be achieved with tool 1 for aluminium and with both tools for copper. The highest variation in the data was noticed for the max. flash heightĥ f (aluminium, tool 1: CV of 30% and copper, tool 2: CV of 17%). Controlling the max. axial forceF z and using surface-coated tools are feasible approaches for more uniform weld surfaces.
Due to different process variables, different measurement methods and a lack of data with regard to the weld topography in literature, a direct comparison to USW and LBW was not possible.

Temperatures in the Weld Surrounding Areas (R5)
The temperatures in the surrounding area of the welds (4.5 mm from the center of the spots, see Appendix B.4) were measured to determine whether thermal damaging of the temperature sensitive separators has to be expected (R5). The data (Figure 5a) showed that all temperatures were far lower than the defined limit of 90 • C. (see Table 2). Figure 5b displays the timeline of one exemplary measurement. It can be seen that the material's temperature rose from room temperature (approx. 20 • C) to a max. valueT (here: 65.47 • C), and quickly declined again. The highest max. valuesT of all experiments were 68.17 • C for aluminium (tool 2) and 65.47 • C for copper (tool 2). The temperatures when using tool 1 were generally lower than when using tool 2, which can be explained by the shorter plunge time (tool 1: 0.24 s, tool 2: 1.56 s, see Section 3.6) and the smaller tool surface area.
the axial force Fz might be suitable for inline control of the weld surfaces.
For copper, the AMs of the max. axial force Fz and the max. spindle torques M z were generally higher than for aluminium, which is due to its higher material strength [31]. Using tool 1, the lowest AM of the max. axial force Fz (1481.19 N) and using tool 2, the lowest SD (28.06 N) and CV (1%) were reached for copper.
The max. spindle torques M z for μFSSW were generally very low (3 to 10 times lower than for FSSW [19]), ranging from 0.44 Nm ± 0.03 Nm (tool 1) to 0.69 Nm ± 0.04 Nm (tool 2) for aluminium and from 0.80 Nm ± 0.04 Nm (tool 2) to 1.02 Nm ± 0.05 Nm (tool 1) for copper with generally very low CV for both tools (aluminium: 6-7%, copper: 5%). No trend or significant correlation could be detected for the max. spindle torque M z.
Based on this analysis, it can be summarized that a high reproducibility (R3) can be achieved with tool 1 for aluminium and with both tools for copper. The highest variation in the data was noticed for the max. flash height ĥf (aluminium, tool 1: CV of 30% and copper, tool 2: CV of 17%). Controlling the max. axial force Fz and using surface-coated tools are feasible approaches for more uniform weld surfaces.
Due to different process variables, different measurement methods and a lack of data with regard to the weld topography in literature, a direct comparison to USW and LBW was not possible.

Temperatures in the Weld Surrounding Areas (R5)
The temperatures in the surrounding area of the welds (4.5 mm from the center of the spots, see Appendix B.4) were measured to determine whether thermal damaging of the temperature sensitive separators has to be expected (R5). The data (Figure 5a) showed that all temperatures were far lower than the defined limit of 90 °C . (see Table 2). Figure  5b displays the timeline of one exemplary measurement. It can be seen that the material's temperature rose from room temperature (approx. 20 °C) to a max. value T (here: 65.47 °C), and quickly declined again. The highest max. values T of all experiments were 68.17 °C for aluminium (tool 2) and 65.47 °C for copper (tool 2). The temperatures when using tool 1 were generally lower than when using tool 2, which can be explained by the shorter plunge time (tool 1: 0.24 s, tool 2: 1.56 s, see Section 3.6) and the smaller tool surface area. It can be concluded that at a distance of 4.5 mm to the spot centers, no thermal damaging of the separators (R5) was to be expected, as the measured temperatures were all lower than the defined limit of 90 °C. It can be concluded that at a distance of 4.5 mm to the spot centers, no thermal damaging of the separators (R5) was to be expected, as the measured temperatures were all lower than the defined limit of 90 • C.
The electrical resistances Rw (R6) of testing samples comprising three spot welds were determined as described in Appendix B5. The testing revealed an increased electrical resistance Rw for all samples compared to unwelded reference foils Rb (Figure 6a), for aluminium from 1413.67 μΩ ± 0.50 μΩ (Rb, Al) to 1577.78 μΩ ± 12.38 μΩ (Rw, Al, tool 2) and for copper from 1334.52 μΩ ± 0.59 μΩ (Rb, Cu) to 1627.53 μΩ ± 0.45 μΩ (Rw, Cu, tool 2). The electrical resistances with tool 1 were higher than with tool 2: 1599.28 μΩ ± 13.82 μΩ (Rw, Al, tool 1) for aluminium, and 1690.57 μΩ ± 2.96 μΩ (Rw, Cu, tool 1) for copper. One explanation for the generally increased electrical resistances Rw is that the foils were only connected through 3 spot welds, which means that only approx. 37% (3 spots with a diameter of 3 mm) of the 24.5 mm wide testing samples were joined. This reduces the cross-sectional area for the electrical current, thereby increasing the resistance in the arrester bar by a factor of approx. 2.70 (inverse of 37%). Also, a slightly higher specific electrical resistance of the arrester material (Cu-ETP) must be added.
Another explanation can be seen in increased conductor lengths inside the arrester bars resulting from the applied testing setup. This thought is schematically shown in Figure 6b. The welds in μFSSW are only created near the surface of the arrester bars (see weld depth). This means that the electrical current flows through the spot welds, through parts of unconnected foils at the lower end of the arrester bars and between those unconnected foils. The hypothesis that decreased weld depths result in higher electrical resistances Rw is supported by the lower electrical resistances produced by tool 2. With that, the theoretical conductor lengths are reduced by approx. 0.55 mm (length of the probe) on both sides of the weld (1.1 mm in total). Using these values and Pouillet's law (see Appendix B.5), the total theoretical increase for the current flow within the unconnected foils can be calculated: 86.39 μΩ for aluminium and 81.55 μΩ for copper. However, the effects in the experimental data are less severe (aluminum: increase of 21.50 μΩ, copper: increase of 63.04 μΩ), which points towards positive influences on the electrical conductivity, such as a larger weld area produced by tool 2 or a higher electrical current flow between the unconnected parts of the foils.
It has to be noted that lower electrical resistances Rw are expected for a different testing setup. The resistance increase in the welding samples was determined simultaneously for both sides of the foil stacks with the setup applied in this study. In practical application, however, the electrical current will flow from the foil via the weld to the closest arrester bar. Additionally, imprecise part fit-up for testing might have resulted in longer measuring lengths, and therefore higher electrical resistances, resulting in errors in the range of several tenths of a μΩ.
From the analysis of the electrical resistances Rw (R6), the use of tool 2 is advised for both materials. Another improvement might be achieved by using tools which enhance One explanation for the generally increased electrical resistances R w is that the foils were only connected through 3 spot welds, which means that only approx. 37% (3 spots with a diameter of 3 mm) of the 24.5 mm wide testing samples were joined. This reduces the cross-sectional area for the electrical current, thereby increasing the resistance in the arrester bar by a factor of approx. 2.70 (inverse of 37%). Also, a slightly higher specific electrical resistance of the arrester material (Cu-ETP) must be added.
Another explanation can be seen in increased conductor lengths inside the arrester bars resulting from the applied testing setup. This thought is schematically shown in Figure 6b. The welds in µFSSW are only created near the surface of the arrester bars (see weld depth). This means that the electrical current flows through the spot welds, through parts of unconnected foils at the lower end of the arrester bars and between those unconnected foils. The hypothesis that decreased weld depths result in higher electrical resistances R w is supported by the lower electrical resistances produced by tool 2. With that, the theoretical conductor lengths are reduced by approx. 0.55 mm (length of the probe) on both sides of the weld (1.1 mm in total). Using these values and Pouillet's law (see Appendix B.5), the total theoretical increase for the current flow within the unconnected foils can be calculated: 86.39 µΩ for aluminium and 81.55 µΩ for copper. However, the effects in the experimental data are less severe (aluminum: increase of 21.50 µΩ, copper: increase of 63.04 µΩ), which points towards positive influences on the electrical conductivity, such as a larger weld area produced by tool 2 or a higher electrical current flow between the unconnected parts of the foils.
It has to be noted that lower electrical resistances R w are expected for a different testing setup. The resistance increase in the welding samples was determined simultaneously for both sides of the foil stacks with the setup applied in this study. In practical application, however, the electrical current will flow from the foil via the weld to the closest arrester bar. Additionally, imprecise part fit-up for testing might have resulted in longer measuring lengths, and therefore higher electrical resistances, resulting in errors in the range of several tenths of a µΩ.
From the analysis of the electrical resistances R w (R6), the use of tool 2 is advised for both materials. Another improvement might be achieved by using tools which enhance the material mixing and result in a higher weld depth (i.e., tools with threads). Further investigations also need to be conducted concerning the positioning and size of the spot welds.

Mechanical Strength (R7)
Tensile testing was performed while the foils were removed from the stack one by one (see Appendix B.6). Almost all aluminium foils failed outside the contacting area (example in Figure 7b) with max. tensile forces F w, Al in the range of the aluminium base material (F b, Al of 33.44 N ± 0.60 N) for both tools (tool 1: 32.11 N ± 1.69 N, CV of 5% and tool 2: welds.

Mechanical Strength (R7)
Tensile testing was performed while the foils were removed from the stack one by one (see Appendix B.6). Almost all aluminium foils failed outside the contacting area (example in Figure 7b) with max. tensile forces Fw, Al in the range of the aluminium base material (Fb, Al of 33.44 N ± 0.60 N) for both tools (tool 1: 32.11 N ± 1.69 N, CV of 5% and tool 2: 31.16 N ± 1.13 N, CV of 4%). This shows that the mechanical strength of the aluminium foils was not impacted by welding. With copper, however, the tensile strengths were lowered due to welding. The foils all failed close to the contacting area (Figure 7c). The reason for the lowered strengths could be the smaller joint areas due to the spot welds (37%, see Section 3.4). This hypothesis is supported by the data from tool 2 (see dotted line in Figure 7). The generally softer material and possible micro force closure outside the spot welds might be the reasons why there was no impact from welding on the strengths of the aluminium samples. The strongest welds for copper were produced by using tool 2 (Fw, Cu, tool 2 of 33.33 N ± 5.07 N, CV of 15%). With tool 1, the max. tensile forces F w, Cu, tool 1 were lower (16.20 N ± 1.90 N, CV of 12%), which might be the result of the reduced weld depths using this tool. Compared to shear testing conducted for LBW joined foil stacks [8], the SD and CV were low. Similarly to the electrical conductivities, it is expected that the strengths in the copper welds can be improved by welding with tools that enhance the material mixing (e.g., structured tools).
It should be noted that the cell design is influenced by the lowest joint strengths. Comparing both materials, it can be concluded that similar strengths (R7) can be achieved using an appropriate tool (tool 1 or tool 2 for aluminium and tool 2 for copper). It should also be noted that a reduction in strength is usual with LBW [8] and USW [2,5].

Welding Speed (R8)
The welding times tweld were calculated (see Appendix B.7) from the process parameters used in the experimental study (plunge speed vp of 25 mm/min, retreat speed vr of 100 mm/min, dwell time td of 0 s and a plunge depth of 0.1 mm, see Appendix A.2). Using tool 1, a welding time tweld, tool 1 of 0.30 s was determined for both materials, which lies in the faster ranges for USW of 40 foils [29]. Using tool 2, the welding time tweld is higher (1.95 s) due to the probe, which had a length of 0.55 mm. With copper, however, the tensile strengths were lowered due to welding. The foils all failed close to the contacting area (Figure 7c). The reason for the lowered strengths could be the smaller joint areas due to the spot welds (37%, see Section 3.4). This hypothesis is supported by the data from tool 2 (see dotted line in Figure 7). The generally softer material and possible micro force closure outside the spot welds might be the reasons why there was no impact from welding on the strengths of the aluminium samples. The strongest welds for copper were produced by using tool 2 (F w, Cu, tool 2 of 33.33 N ± 5.07 N, CV of 15%). With tool 1, the max. tensile forces F w, Cu, tool 1 were lower (16.20 N ± 1.90 N, CV of 12%), which might be the result of the reduced weld depths using this tool. Compared to shear testing conducted for LBW joined foil stacks [8], the SD and CV were low. Similarly to the electrical conductivities, it is expected that the strengths in the copper welds can be improved by welding with tools that enhance the material mixing (e.g., structured tools).
It should be noted that the cell design is influenced by the lowest joint strengths. Comparing both materials, it can be concluded that similar strengths (R7) can be achieved using an appropriate tool (tool 1 or tool 2 for aluminium and tool 2 for copper). It should also be noted that a reduction in strength is usual with LBW [8] and USW [2,5].

Welding Speed (R8)
The welding times t weld were calculated (see Appendix B.7) from the process parameters used in the experimental study (plunge speed v p of 25 mm/min, retreat speed v r of 100 mm/min, dwell time t d of 0 s and a plunge depth of 0.1 mm, see Appendix A.2). Using tool 1, a welding time t weld, tool 1 of 0.30 s was determined for both materials, which lies in the faster ranges for USW of 40 foils [29]. Using tool 2, the welding time t weld is higher (1.95 s) due to the probe, which had a length of 0.55 mm.
The use of multi-spindle heads is suggested for a production of several spots at once. Since the process parameters for this study were not chosen with the objective of a minimized production time, lower welding times are expected to be possible.

Conclusions and Outlook
Through an experimental study, it was shown that µFSSW is a promising new joining technology for internal contacting in prismatic lithium-ion battery cells. Spot welds were fabricated for stacks consisting of 30 aluminium and 30 copper foils with two different tools (with and without a probe). The process produced welds with good properties, with regard to pre-defined requirements (R1-R8) taken from literature. The most important findings are:

•
Using µFSSW, sound aluminium and copper spot welds without visible defects or ruptured foils can be produced. Due to the low temperatures, no thermal damaging of the battery components has to be expected.

•
Aside from the multi-sheet lap joint configuration [21,32], the process also allows cell-internal contacting in multi-sheet butt joint configuration, which increases the flexibility in cell design.

•
The tensile strengths of the aluminium samples were not impacted by welding. For the copper samples, the strengths were reduced to that of aluminium. It is hypothesized that the mechanical quality can be improved by welding with tools that enhance the material mixing (e.g., structured tools).

•
For both aluminium and copper, the electrical resistances of the welding samples were higher than that of unwelded reference foils. It is expected that resistances can be improved by increasing the cross-section area of the weld, for example, by welding with larger or with structured tools. Another possibility is to produce line welds [32]. • Spot welds can be produced in 0.3 s. The welding speed can be improved by increasing the plunge (v p ) and the retreat speed (v r ) of the tool.

•
In addition, a scaled up prismatic demonstrator, which is comprised of 100 uncoated aluminium and 100 uncoated copper foils, was produced with two arrester bars on each side ( Figure 8) to show the potential of the process for the production of large-scale high-energy-dense battery cells.

•
Like USW [2], µFSSW [33] is a solid-state welding process with a low energy consumption and no fumes (in contrast to LBW). The welding times of µFSSW, USW, and LBW are similar, however, with µFSSW, the required welding times are not increased by higher foil stack sizes. Both multi-sheet lap joints and multi-sheet butt joints can be fabricated by LBW and by µFSSW, but not by USW. LBW and µFSSW therefore allow for more flexibility with regard to the cell design. Both processes, however, also require good joint fit-up [2], which means a higher effort for positioning and clamping before welding.

•
The experimental results show that µFSSW has a high potential for cell-internal contacting of prismatic cells. An application for other cell formats like cylindrical cells is regarded feasible by the authors and might be advantageous due to the high uniformity of the spot welds and the high speed of the process. • Due to different applied materials and welding geometries, a literature-based comparison between µFSSW, USW, and LBW was not possible for all defined requirements. This will be a topic of future investigations. • Future research will also deal with an assertion of the electrochemical performance of lithium-ion cells, which were produced using µFSSW. The use of multi-spindle heads is suggested for a production of several spots at once. Since the process parameters for this study were not chosen with the objective of a minimized production time, lower welding times are expected to be possible.

Conclusions and Outlook
Through an experimental study, it was shown that μFSSW is a promising new joining technology for internal contacting in prismatic lithium-ion battery cells. Spot welds were fabricated for stacks consisting of 30 aluminium and 30 copper foils with two different tools (with and without a probe). The process produced welds with good properties, with regard to pre-defined requirements (R1-R8) taken from literature. The most important findings are:

•
Using μFSSW, sound aluminium and copper spot welds without visible defects or ruptured foils can be produced. Due to the low temperatures, no thermal damaging of the battery components has to be expected.

•
Aside from the multi-sheet lap joint configuration [21,32], the process also allows cellinternal contacting in multi-sheet butt joint configuration, which increases the flexibility in cell design.

•
The tensile strengths of the aluminium samples were not impacted by welding. For the copper samples, the strengths were reduced to that of aluminium. It is hypothesized that the mechanical quality can be improved by welding with tools that enhance the material mixing (e.g., structured tools).

•
For both aluminium and copper, the electrical resistances of the welding samples were higher than that of unwelded reference foils. It is expected that resistances can be improved by increasing the cross-section area of the weld, for example, by welding with larger or with structured tools. Another possibility is to produce line welds [32]. • Spot welds can be produced in 0.3 s. The welding speed can be improved by increasing the plunge (vp) and the retreat speed (vr) of the tool.

•
In addition, a scaled up prismatic demonstrator, which is comprised of 100 uncoated aluminium and 100 uncoated copper foils, was produced with two arrester bars on each side (Figure 8) to show the potential of the process for the production of largescale high-energy-dense battery cells. • Like USW [2], μFSSW [33] is a solid-state welding process with a low energy consumption and no fumes (in contrast to LBW). The welding times of μFSSW, USW, and LBW are similar, however, with μFSSW, the required welding times are not

Acknowledgments:
We thank the AWM Maschinenbau GmbH for the fabrication of the clamping system and the Hufschmied Zerspanungssysteme GmbH for supplying the welding tools.

Conflicts of Interest:
The authors declare that they have no conflict of interest.

Appendix A. Setup for the Welding Experiments
Appendix A.1. Welding Machine, Clamping System and Part Fit-Up The welding experiments were conducted in position-controlled mode using a fouraxis horizontal milling machine (MCH 250, Gebr. Heller Maschinenfabrik GmbH, Nuertingen, Germany), which had been adapted for friction stir welding.
A stainless-steel clamping system ( Figure A1) was designed for the welding task, which consisted of two clamping shoes, each with a groove for one arrester bar. The shoes were mounted on a vise for automatic and fast clamping using air pressure. Two additional fixture plates with comb-shaped openings for access of the welding tool were pressed against the bar surfaces and prevented misalignment. For fit-up, the arrester bars were first placed inside the grooves. Afterwards, the foil stack was positioned between the shoes, so that the edges of the foils were flush with the arrester bar surfaces. After clamping the foil stack and the two arrester bars, the combshaped plates were mounted and fixed with screws.

Appendix A.2. Welding Tools and Process Parameters
A sketch of the tools used for experimentation is shown in Figure A2. Two tools were designed: one without a probe (tool 1, displayed in Figure A2a) and one with a probe (tool 2, displayed in Figure A2b). Both tools had a flat and unstructured shoulder with a diameter of 3 mm. For fit-up, the arrester bars were first placed inside the grooves. Afterwards, the foil stack was positioned between the shoes, so that the edges of the foils were flush with the arrester bar surfaces. After clamping the foil stack and the two arrester bars, the comb-shaped plates were mounted and fixed with screws.
GmbH & Co. KG, Ulm, Germany) with a 1 kN load cell (Xforce HP, ZwickRoell GmbH & Co. KG, Ulm, Germany). Due to the shape of the weld geometries, no standard tensile testing could be performed. Instead, the arrester bars were loosely fixed, so that they did not compress the foil stack during testing ( Figure A4). Single foils (the 7th, the 15th, and the 23rd foil) were tested one by one to determine the bonding quality. A testing speed of 50 mm/min was applied. Figure A4. Sketch of the weld testing sample for mechanical testing; geometrical values are not to scale and are given in mm if not otherwise indicated.

Appendix B.7. Welding Time Calculation
The welding time tweld was calculated from the process parameters plunge speed vp, retreat speed vr, plunge depth dp, and dwell time td: tweld = dp/vp + td + dp/vr.
It should be noted that the welding time tweld does not include the time for part fitup, clamping, and unclamping, since no comparable data for these steps were found in literature for USW and LBW.  It should be noted that the welding time t weld does not include the time for part fit-up, clamping, and unclamping, since no comparable data for these steps were found in literature for USW and LBW.