Journal of Power Sources 272 (2014) 356e361
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
Laser additive manufacturing of stainless steel micro fuel cells Gianmario Scotti a, *, Ville Matilainen b, Petri Kanninen c, Heidi Piili b, Antti Salminen b, d, Tanja Kallio c, Sami Franssila a a
Department of Materials Science and Engineering, Aalto University, P.O. Box 16200, 00076 Aalto, Finland Laser Processing Research Group, Lappeenranta University of Technology, Tuotantokatu 2, 53850 LPR Lappeenranta, Finland c Department of Chemistry, Aalto University, P.O. Box 16100, 00076 Aalto, Finland d €isenkatu 28, 20520 Turku, Finland Machine Technology Centre Turku Ltd., Lemminka b
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
First ever stainless steel micro fuel cells made by laser additive manufacturing. Maximum current density: 1.19 A cm2, maximum power density: 238 mW cm2. The results are comparable to those of macro fuel cells. We demonstrated that the method is suitable for fast prototyping. The method was used to test flowfield aspect ratio modifications.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 21 June 2014 Received in revised form 29 July 2014 Accepted 20 August 2014 Available online 2 September 2014
This paper introduces laser additive manufacturing as a new method for the fabrication of micro fuel cells: The method opens up the capability of ultrafast prototyping, as the whole device can be produced at once, starting from a digital 3D model. In fact, many different devices can be produced at once, which is useful for the comparison of competing designs. The micro fuel cells are made of stainless steel, so they are very robust, thermally and chemically inert and long-lasting. This enables the researcher to perform a large number of experiments on the same cell without physical or chemical degradation. To demonstrate the validity of our method, we have produced three versions of a micro fuel cell with square pillar flowfield. All three have produced high current and power density, with maximum values of 1.2 A cm2 for the current and 238 mW cm2 for power. © 2014 Elsevier B.V. All rights reserved.
Keywords: 3D printing Proton exchange membrane Stainless steel Fuel cells Prototyping
1. Introduction Li-ion secondary cell batteries are the current power source of choice for portable electronics, but micro fuel cells (MFCs) have the potential to further increase energy density [1,2]. MFCs have most commonly been microfabricated from silicon [3e9] but other
* Corresponding author. Tel.: þ358 503632739. E-mail address:
[email protected] (G. Scotti). http://dx.doi.org/10.1016/j.jpowsour.2014.08.096 0378-7753/© 2014 Elsevier B.V. All rights reserved.
materials have been utilized, such as metalized PMMA [10e13], SU8 [14], PDMS [15] [16], pyrolized carbon [17], bulk aluminum [18] and 50 mm thin hydroformed stainless steel sheets [19]. These materials and techniques do not allow for easy disassembly and reassembly of the cells, and typically, replacing the membraneelectrode assembly (MEA) or the gas diffusion layer (GDL) requires also the replacement of the bipolar plates. Furthermore, few techniques are suitable for fast prototyping: a typical microfabrication process flow requires the preparation of one or more photomasks, the utilization of one or more photolithography steps,
G. Scotti et al. / Journal of Power Sources 272 (2014) 356e361
and processing with separate tools for the etching of the microchannels and for the deposition of a metallic current collector. While moulding, stamping, and hydroforming are faster methods once the mould, or die is ready, preparation of the mould is, again, a slow process. Among the established techniques only laser ablation [10,13,20] and CNC machining [11,12] are suitable for rapid prototyping, in the sense that a testable MFC can be produced directly from a design made on a computer, with few processing steps. In our previous work we have optimized the laser ablation parameters for speed [20], disregarding the small irregularities in flow-field channels, as they proved to be inconsequential to the functioning of the MFCs. However, even with the high processing speeds achieved by methods such as laser ablation or CNC machining, the drawbacks of the serial subtractive approach come to the fore when large cavities must be produced. Many MFC designs require the presence of large volume cavities, such as fuel reservoirs [3,4,11], or basins to accommodate thick gas diffusion layers [7,18]. Speeding up CNC machining by using larger milling bits with CNC, or increasing laser ablation speed with larger beam waist and sustained power, is not usually feasible, as these actions would then make it impossible to create sufficiently small microchannels or inlet holes. Finally, it should be mentioned that CNC-milled [11,12] or laser-ablated [10,13] PMMA flow-field plates require a separate metallization step, so that they may collect the produced current. In this work we introduce laser additive manufacturing (LAM) for the rapid prototyping of stainless steel micro-fuel cells: LAM is a layer-wise material addition technique where complex 3D parts are manufactured by selective melting and solidification of consecutive layers of powder material on top of each other (Fig. 1) [21,22]. LAM can be used both for rapid prototyping as well as for manufacturing of complex metallic objects. The material for our MFCs is 316L stainless steel, a corrosion and acid resistant steel alloy also used for medical implants. Its excellent corrosion resistance combined with its high hardness and toughness permits the experimenter to reuse the MFCs fabricated from it, many times. The electrical resistivity of 316L stainless steel is 74 mU cm, which compares favorably to the 10 mU cm of highly-doped silicon used in Refs. [7,20]. Because of the low electrical resistivity, the steel flowfield plates are good current collectors.
357
To assess the viability of LAM for MFC fast prototyping, we have fabricated and tested three variants of a cell with square pillar flowfield and inserted carbon cloth GLD. The difference between the three designs was the size and aspect ratio of the flowfield: 1 1 cm2, 2 1 cm2, and 4 1 cm2. Scaling the flowfield in one dimension is intended to determine if elongated MFCs, suitable for form factors of devices such as mobile phones, would still generate sufficient power and current density compared with 1 1 aspect ratio flowfields. 2. Experimental 2.1. MFC structure and construction The design of the MFCs in this work is similar to the one in Refs. [7,18]: we have a square pillar flowfield and a basin for accommodating a commercial carbon cloth GDL. An exploded schematic view of the MFC construction is presented on Fig. 2a. All the main feature sizes are summarized on Fig. 2b: the pillars have a 1 mm 1 mm cross section, while the channel (inter-pillar distance) is 500 mm wide. There is a 200 mm gap from the top of the pillars to the edge of the flowfield plate. This is done in order to accommodate a commercial carbon cloth GDL. The GDL used was GDL-CT® (Fuel Cell Etc), with the microporous side turned towards the MEA. The GDL was 410 mm thick when not compressed. The MEA was a Gore® Primea membrane (a proton-conductive ionomer similar to Nafion®) with a platinum loading of 0.1 mg cm2 on the anode and 0.3 mg cm2 on the cathode side. Thread seal Teflon® tape (also known as plumber's tape) was stretched on the edge of the flowfield plates, to act as a simple gasket. The MFC was mounted in a matching jig made of ABS polymer by 3D printing. Fig. 3 shows the 3D models of the three MFC flowfield plates fabricated by LAM. These plates differ only by the flowfield size in one dimension. The flowfield sizes are (a) 1 cm 1 cm, (b) 2 cm 1 cm, and (c) 4 cm 1 cm. The MFCs made with these plates will from now on be called “1 1”, “2 1”, and “4 1”, for convenience. 2.2. LAM of flowfield plates Fig. 1 presents the process cycle of LAM. In this process, we first create a digital 3D model of the object to be manufactured. This 3D
Fig. 1. Diagram of the laser additive manufacturing (LAM) process. On the left: file manipulation from 3D solid model to 3D STL model and to 2D slices. Center: basic principle of laser additive manufacturing process. First the powder is spread then the laser beam melts the geometry of one layer. Finally, then building platform is lowered. The powder is then spread again, and the cycles repeat until the part is finished.
358
G. Scotti et al. / Journal of Power Sources 272 (2014) 356e361
Fig. 2. Construction of an MFC with LAM-fabricated flowfield plates. (a) Exploded view. (b) Cross section of a flowfield plate with annotated feature sizes. The top of the pillars is 200 mm below the edge of the flowfield, to accommodate the GDL.
model is then sliced in the Z direction into 20 mm thick layers. Using these layers, a cross-section of desired geometry is melted locally by a laser beam to form a solid layer. Once the cross-section is melted a new layer of powder material is spread. This cycle is repeated until the desired object is finished. The platform on which the pieces rest is preheated to 80 C, to decreases temperature gradient and internal stresses during the manufacturing. The process usually takes place in an inert atmosphere, such as nitrogen or argon, to avoid oxidation [21,22]. In our case the chamber was filled with 99.8% pure nitrogen gas. Once the building process is finished the part is surrounded by loose powder, which is removed and it can be reused after sieving. The part itself must be detached from the platform, and usually
needs some post-processing. In our case the flowfield plates were placed with the longer edge on top of the platform for easier detachment, and to make the best use of the available real estate. The powder used for LAM of the MFC flowfield plates was EOS StainlessSteel 316L, which is a 316L stainless steel with median particle size of 31 mm. The main components of the alloy, by weight %, are Fe 62%, Cr 17e19%, Ni 13e15%, Mo 2.25e3%, Mn 2%, Si 0.75%, and Cu 0.5%. Because of the relatively large particle size of the powder, the edge of the plate on which the Teflon® tape was stretched required some polishing, which was done with a wet grinder. This was necessary to ensure good gas sealing. The LAM equipment used in the experiment can be seen on Fig. 4: the bespoke setup was assembled by EOS GmbH, and it uses a 200 W continuous wave yttrium fiber laser source operating at 1070 nm wavelength, and a Scanlab hurrySCAN® 20 galvanometer scanner (visible on the top of the process chamber) for fast beam movement.
Fig. 3. Three versions of LAM-fabricated flowfield plates. The flowfield sizes are (a) 1 cm 1 cm, (b) 1 cm 2 cm, and (c) 1 cm 4 cm. The inset shows an enlarged view of the model near one of the inlet holes.
Fig. 4. Photograph of the LAM setup at Lappeenranta University of Technology. On the right side are the control computer and the laser source. In front is the process chamber, with the galvanometric scanner on top.
G. Scotti et al. / Journal of Power Sources 272 (2014) 356e361
The processing time for 10 pieces of 1 1 flowfield plates (2 mm 14 mm 14 mm total volume for one 1 1 plate, Fig. 2) is 5 h, and if the whole building platform were utilized with 154 pieces of 1 1 plates the building time would be ~30 h. These building times are simulated by the LAM machine's process control software PSW 3.2. In our case, we fabricated four of each type of flowfield plate (Fig. 3) which took about 6 h 30 min. 2.3. MFC characterization The MFCs were fueled with humidified hydrogen, while the oxidant was non-humidified oxygen. The cells were not heated, so their temperature was near room temperature of 20 C. The flow rate of both gases was kept at 25 mL min1 per cm2 of flowfield area. This means that the gases were flowing at a rate of 25 mL min1 in the 1 1 cells, 50 mL min1 in the 2 1 cells and 100 mL min1 in the 4 1 cells. The over-pressure was estimated at ~0.1 bar. Polarization and chronoamperometric measurements were both performed using a Metrohm Autolab PGSTAT100 potentiostat. For the polarization curve measurements a BSTR10A booster was added to the setup, because the total current produced by the micro fuel cells exceeded the capacity of the potentiostat alone. The experimental protocol was as thus: 1) First a set of polarization curves were measured. The measurement was repeated until the curves became stable and did not differ between iterations. The polarization curves were obtained by sweeping the potential across the cells at a rate of 3 mV s1, from open-circuit potential of about 0.97 V down to 0.05 V. 2) Once the fuel cell performance was deemed stable, a chronoamperometric test was performed, for 15 h. During the chronoamperometric measurement the potential on the cells was kept at 0.6 V. 3) After the chronoamperometric experiment was over, a new set of polarization curves was obtained, much the same way as in step 1). An additional experiment was made, with the 1 1 MFC: polarization curves were obtained with the gas flow rates increased from 25 mL min1 to 50 mL min1, which are the same conditions as in Refs. [7,18]. This allowed us to make a better comparison with published works. 3. Results and discussion For the 1 1 cell, a maximum current density of 1172 mA cm2 and a power density of 233 mW cm2 were obtained. When the gas flow rates of the 1 1 cell were increased from 25 mL min1 to 50 mL min1 to better compare the steel MFCs with our previous results [18], the maximum power density increased further to 238 mW cm2 and the current density to 1190 mA cm2. This power density is nearly double compared to the silicon MFCs in Ref. [7] (127 mW cm2) and an improvement compared to the aluminum MFCs in Ref. [18] (228 mW cm2). These values are comparable even with macro fuel cells made of 316 stainless steel (similar to the 316L used in this work) [23]. Note that in Ref. [23] the cells were kept at 50 C and there was an overpressure of 2 bar for both gases, whereas our set-up was run under conditions normal for micro fuel cells: very small over-pressure and close to room temperature (~20 C). Table 1 contains a summary of the MFC characterization results, while. Figs. 5e7 show all the obtained results, using the experimental protocol outlined in Subsection 2.3. The maximum current
359
Table 1 Summary of MFC characterization results. The results in the 2nd and 3rd column relate to the polarization curves made after the chronoamperometric measurements. The 4th column contains an averaged value of the current density obtained during the time interval between 2 h and 15 h of the chronoamperometric measurement. MFC type
Max. current density [mA cm2]
Max. power density [mW cm2]
Average sustained current density (2 he15 h) [mA cm2]
11 21 41
1172 1089 848
233 217 164
214 205 174
density for the 2 1 MFC was 1089 mA cm2 and the power density was 217 mW cm2, while for the 4 1 cell the values were 848 mA cm2 for current and 164 mW cm2 for power density. Both of these results are still comparable or better than the best MFCs in literature. Comparing the MFCs with each other, we see that the longer the flowfield, the lower the maximum current and power densities (Figs. 5 and 6). Also the sustained current density decreases with the longer flowfields (Fig. 7). This is likely due to a lower and more uneven contact pressure between the components caused by clamping the MFC by bolts [24]. However, the loss in performance is relatively small d less than 10% going from 1 1 to 2 1 and 15% from 2 1 to 4 1 for the sustained current density (Table 1) d and if these cells were to be used in a portable electronic device, it would be rational to adjust the form factor of the MFC to that of the device it powers; we will obtain higher power and overall higher energy density by using a 4 1 cell instead of three 1 1 cells. It was observed that the performance of all the cells presented in this work was in general excellent. This can be attributed to the presence of the thick GDL integrated with the MFC flowfield. The same approach was used in Refs. [7,18], where it was very beneficial. Fig. 8 is a photograph of the LAM fabricated flowfield plates with a 1 EUR coin for size comparison. Fig. 9 shows an optical micrograph of the LAM flowfield plate. The rough surface of the flowfield, due to the relatively large median size of the steel powder particles (31 mm) that composes it, increases the electrical contact surface between the flowfield plates and the GDL. This, also, was observed in Refs. [7,18], although the nature of the asperity in those previous studies was entirely different from the one present with LAM fabricated flowfield plates: in Ref. [7] we used silicon nanograss for contact enhancement, in Ref. [18] the asperity was caused by non-etched components in the aluminum alloy, and in this work, the asperity is the result of the particle size of the sintered steel dust. From Fig. 9 we can draw the conclusion that the dimensions of the pillars are not precisely 1 mm 1 mm. This could be an issue with more complex flowfield topologies, especially with channels narrower than 500 mm. According to our measurements, the 200 mm gap from the top of the pillars to the upper edge of the flowfield plates was reproduced very accurately (Fig. 2b), guaranteeing an even compression force across the GDL. All three MFC designs showed improved performance after the chronoamperometric measurement due to catalyst activation [25] seen as an earlier increase of current between 1.0 V and 0.8 V in Fig. 5b compared to Fig. 5a. As mentioned previously, fabrication of four of each type of flowfield plate took approximately 6 h 30 min. In Ref. [26], flowfield plates from 316L stainless steel were prepared by CNC micromilling. The authors report a milling time of 5 min for a 47.5 mm long, 300 mm deep and 1 mm wide channel, which results in a milling speed of 2.85 mm3 per minute. Since the total
360
G. Scotti et al. / Journal of Power Sources 272 (2014) 356e361
Fig. 5. Polarization curves showing the current densities obtained (a) before, and (b) after the 15 h chronoamperometric measurement. The current density, for all three cell designs, has increased after the polarization measurement.
Fig. 6. Polarization curves showing the power density (a) before, and (b) after the 15 h chronoamperometric measurement. Power density has increased after the 15 h chronoamperometric measurement, for all three MFC designs.
Fig. 7. Summary of the 15 h chronoamperometric measurement results. The voltage across the cells was kept at 0.6 V. After a brief fall, the current density kept increasing in all three MFC designs.
Fig. 8. Photograph of the LAM fabricated flowfield plates next to a 1 EUR coin, for size comparison. The flowfield edge is somewhat irregular, but that does not present a problem for the intended use.
G. Scotti et al. / Journal of Power Sources 272 (2014) 356e361
361
Fig. 9. Low magnification micrographs of LAM fabricated flowfield plate. (a) Optical micrograph of slanted view. (b) SEM top view of an inlet hole.
volume of the channels, the inlet holes, and the basin in our 1 1 type cell is ~27.5 mm3, the fabrication time using CNC milling would have been close to 10 min. The total time for 4 of each type of flowfield plate would have been 4 h 30 min, not counting the time to dice/separate the CNC milled plates from the stainless steel substrate. This compares favorably to LAM. However, with deeper channels and larger cavities LAM has a speed advantage over CNC milling, where processing time increases with removed volume. 4. Conclusions In this work we have validated the method of using LAM for rapid prototyping of MFCs: we found that the whole process from the CAD design of three different versions of an MFC, to finished, testable prototypes, is straightforward and quick. We also found that the performance of the MFCs was excellent for all three designs. Finally, the study yielded a useful finding on the aspect ratio of MFC flowfields: we came to the conclusion that the performance of the MFC can scale almost linearly with the lengthening of the flowfield. This is a valuable result since the intended purpose of MFCs is to power mobile devices, which can have widely varying form factors. The large size of the steel powder used in LAM of the flowfield plates creates a rough surface, which is not necessarily bad; such surface increases the contact between the current collecting flowfield and the GDL. We also discovered that the feature sizes are not controlled very accurately, at the sub-millimeter scale of the MFCs in this work. Considering the high performance obtained in this work, this size variation does not seem to matter for a square pillar flowfield with a thick GDL on top of it. However, other types of MFCs and flowfields may require better dimensional control, which should be possible to achieve by using powders with smaller particle sizes, and by optimizing the geometry of the manufactured cells. For instance, Sandvik Osprey Ltd. produces 316L stainless steel powder with 5 mm particle size. The geometry optimization can be done so that the MFCs better follow the design rules of additive manufacturing. For example, there should not be structures just laying on top of powder since the powder does not serve as support and the minimum build angle for stainless steel material is 45 . Also the build orientation has an effect on the quality and accuracy of the features to be manufactured. For instance, by optimizing the build orientation, the overhanging features can be minimized and with this optimization the pillars could be more accurate.
Acknowledgments The MFC flowfield plates were fabricated at Lappeenranta University of Technology. The MFC characterization was done at the Fuel Cell Laboratory, Department of Chemistry, Aalto University.
References [1] A. Kundu, J.H. Jang, J.H. Gil, C.R. Jung, H.R. Lee, S.H. Kim, B. Ku, Y.S. Oh, J. Power Sources 170 (2007) 67e78. [2] E. Kjeang, N. Djilali, D. Sinton, J. Power Sources 186 (2009) 353e369. [3] S. Aravamudhan, A.R.A. Rahman, S. Bhansali, Sens. Actuators A 123e124 (2005) 497e504. [4] J. Yeom, R.S. Jayashree, C. Rastogi, M.A. Shannon, P.J.A. Kenis, J. Power Sources 160 (2006) 1058e1064. [5] S. Sundarrajan, S.I. Allakhverdiev, S. Ramakrishna, Int. J. Hydrogen Energy 37 (2012) 8765e8786. [6] S. Kouassi, G. Gautier, J. Thery, S. Desplobain, M. Borella, L. Ventura, J.Y. Lautent, J. Power Sources 216 (2012) 15e21. [7] G. Scotti, P. Kanninen, T. Kallio, S. Franssila, J. Micromech. Microeng 22 (2012) 094006. [8] B. Scherrer, et al., J. Power Sources 258 (2014) 434e440. [9] X. Wang, Y. Zhou, Q. Zhang, Y. Zhu, L. Liu, J. Micromech. Microeng 19 (2009) 094012. [10] S.-S. Hsieh, J.-K. Kuo, C.-F. Hwang, H.-H. Tsai, Microsyst. Technol. 10 (2004) 121e126. [11] N. Hashim, S.K. Kamarudin, W.R.W. Daud, Int. J. Hydrogen Energy 34 (2009) 8263e8269. [12] J.A. Alanis-Navarro, C. Reyes-Betanzo, J. Moreira, P.J. Sebastian, J. Power Sources 242 (2013) 1e6. [13] S.H. Chan, N.-T. Nguyen, Z. Xia, Z. Wu, J. Micromech. Microeng 15 (2005) 231e236. ndez, J. Santander, M. Lo €rgen, [14] J.P. Esquivel, T. Senn, P. Hern andez-Ferna € chel, C. Cane , N. Sabate , J. Power Sources 195 (2010) B110eB115. S. Rojas, B. Lo [15] K. Shah, W.C. Shin, R.S. Besser, Sens. Actuators B 97 (2004) 157e167. [16] Y.-A. Song, C. Batista, R. Sarpeshkar, J. Han, J. Power Sources 183 (2008) 674e677. [17] Y. Wang, L. Pham, G.P.S. de Vasconcellos, M. Madou, J. Power Sources 195 (2010) 4796e4803. [18] G. Scotti, P. Kanninen, T. Kallio, S. Franssila, J. Microelectromech. Syst. 23 (2014) 372e379. [19] M. Koc, S. Mahabunphachai, J. Power Sources 172 (2007) 725e733. [20] G. Scotti, D. Trusheim, P. Kanninen, D. Naumenko, M. Schulz-Ruhtenberg, V. Snitka, T. Kallio, S. Franssila, J. Micromech. Microeng 23 (2013) 055021. [21] V.-P. Matilainen, Characterization of Process Efficiency Improvement in Laser Additive Manufacturing, Lappeenranta University of Technology, 2014. Master's thesis BK10A0401. [22] H. Schleifenbaum, W. Meiners, K. Wissenbach, C. Hinke, CIRP J. Manuf. Sci. Technol. 2 (2010) 161e169. [23] D.P. Davies, P.L. Adcock, M. Turpin, S.J. Rowen, J. Power Sources 86 (2000) 237e242. [24] M.S. Opu, D. Choi, M. Ohashi, S. Shimpalee, J.W. Van Zee, ECS Trans. 33 (2010) 1017e1025. [25] Z. Qi, A. Kaufman, J. Power Sources 111 (2002) 181e184. [26] H.S. Jang, D.S. Park, Sensors 10 (2010) 167e175.