Development of an All Kapton-Based Thin-Film ... - IEEE Xplore

IEEE SENSORS JOURNAL, VOL. 14, NO. 10, OCTOBER 2014

3377

Development of an All Kapton-Based Thin-Film Thermocouple Matrix for In Situ Temperature Measurement in a Lithium Ion Pouch Cell Nora Martiny, Member, IEEE, Alexander Rheinfeld, Jan Geder, Yuxi Wang, Werner Kraus, and Andreas Jossen

Abstract— Based on the design of a lithium ion battery cell and the resulting thermophysical properties, considerable temperature gradients may form within the cell not only during abusive scenarios. While the temperature gradients during normal operation are mostly negligible for small cells, which are commonly employed in mobile applications, the discrepancy between the temperature inside a battery cell and the cell’s surface can be significant for larger scaled cells, which are employed, e.g., in automotive applications. Battery management systems that rely on a monitoring of the cell’s surface temperature may consequently lead to an unfavorable operation of the whole battery in terms of aging and safety aspects. Thus, we hereby present an approach of designing, producing, and incorporating an all Kapton-based temperature sensor for an in situ temperature monitoring of lithium ion pouch cells. First prototypes are developed as a proof of concept when using a common potential for a thermocouple matrix that will later allow for a spatially resolved temperature monitoring within a lithium ion pouch cell. The interactions between the sensor and the cell are investigated and further design steps for improvements in functionality are elaborated. Index Terms— Lithium ion batteries, temperature monitoring, thin-film, Kapton, thermocouple, in-situ, sensor matrix, common potential, pouch cell, temperature sensor.

I. I NTRODUCTION ITHIUM ion batteries are the first choice for most of today’s applications including power tools, mobile devices and electric vehicles, thanks to their high energy and power density, high voltage, enhanced lifetime and the absence of severe memory effects compared to batteries with other chemistries. Still, some challenges need to be faced which are mainly related to aging and safety. While fast aging mainly results in higher costs for replacement, safety aspects

L

Manuscript received March 31, 2014; revised May 7, 2014; accepted June 17, 2014. Date of publication June 19, 2014; date of current version August 18, 2014. This work was supported by the Singapore National Research Foundation through the Campus for Research Excellence and Technological Enterprise Program. This is an expanded paper from the IEEE SENSORS 2013 Conference. The associate editor coordinating the review of this paper and approving it for publication was Prof. David A. Horsley. N. Martiny, J. Geder, and Y. Wang are with TUM CREATE Ltd., Singapore 138602 (e-mail: [email protected]; jan.geder@ tum-create.edu.sg; [email protected]). A. Rheinfeld and A. Jossen are with the Institute for Electrical Energy Storage Technology, Technical University of Munich, Munich 80333, Germany (e-mail: [email protected]; [email protected]). W. Kraus is with the Institute for Technical Electronics, Technical University of Munich, Munich 80333, Germany (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JSEN.2014.2331996

are clearly more critical. Due to the relatively high voltage level, organic electrolytes need to be employed in order to ensure a meta-stability during operation via forming protective layers on the surface of the electrodes. Potential damages to these protective layers (e.g. through mechanical influences) combined with a comparably high level of flammability of the applied electrolyte inevitably result in safety hazards. While higher overall temperatures in the cell mainly lead to an accelerated aging (that is a loss of capacity and a rise in internal resistance), an internal cell temperature exceeding the designated temperature range during cycling may be an early indicator for safety relevant failures. With the overall trend of increasing cell sizes, especially in automative applications, considerable inhomogenities within the temperature distribution need to be faced. To find optimal operating conditions and for early failure detection, permanent, space resolved and reliable temperature monitoring is therefore crucial. In most of today’s applications the monitoring is performed by attaching sensors to the surface of the cells (see [1]). However, active cooling systems are frequently used in larger scale applications which are cooling the battery from its surface. The resulting difference in time response between a cell’s surface and its interior may result in an inappropriate battery management when only the surface temperature is monitored. Therefore, an in-situ temperature monitoring as it is described by [2]–[4] is advantageous. Furthermore, reactions in the worst case situation of a thermal runaway can happen fast and may not be detectable on the cell’s surface within a timescale that allows to shut off the cell before serious damage to the surroundings occurs. Feng et al. show in [5] that within 10 ms the internal temperature can rise by more than 200 K. A sensor placed in a 25 Ah prismatic hard case cell which is formed of two pouch cells showed considerably large temperature gradients between the cell’s center and the cell’s surface (> 500 K) during a thermal runaway test. Previous research activities presented by [6] showed that tracking the in-cell temperature is mainly beneficial during abuse scenarios and that a space resolved temperature monitoring is crucial in order to detect critical hotspots and undesired temperature gradients. Referring to thermographic data of a 20 Ah pouch cell for varying discharge rates the cell internal temperature is estimated via simulation. In this paper a novel approach is presented for the development of a thin-film temperature sensor matrix that can be

1530-437X © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

3378


placed directly into a lithium ion pouch cell which provides space resolved data from there. The remainder of this paper describes design criteria that need to be taken into account and the resulting sensor design and placement in Section II. In Section III the sensor production process is elaborated, followed by a description of the conducted experiments and their results in Section IV. The paper is concluded in Section V. II. S ENSOR D EVELOPMENT The overall criterion that is to be considered for in-situ temperature sensors is to affect the behavior of the battery cell as little as possible. To achieve this, different requirements for the sensor design are set. A. Shape Requirements Wire-based sensors, as described in [3] or [7], showed leakage at the sealing of the aluminum laminate casing of the pouch cell after a few days. Furthermore it is assumed that a flat film sensor reduces the mechanical stress within the electrodes compared to a wire-based sensor, as the electrodes are not bent in the sensor area during vacuum sealing. A further advantage of a thin-film design is that the thermal impact when including such a sensor is limited due to the relatively small thermal resistance and the thermal mass of the temperature sensor. Examples for such a thin-film temperature sensor with similar design requirements are given by [8]–[10] for fuel cell applications and by [2] for batteries. To gather space resolved temperature data for better detection of local inhomogeneities, a sensor matrix with measurement points distributed over the whole area of the cell is desirable. A simple solution that allows small measurement areas and therefore more measurement points within the area is a thermocouple matrix with a single common potential (see Fig. 2). B. Chemical Requirements In state-of-the-art lithium ion cells, LiPF6 -based electrolyte solutions are widely used. However, traces of water, moisture or alcohols can react with the decomposition components of LiPF6 according to (1) and (2) [11] and small amounts of hydrofluoric acid (HF) can arise. Therefore, stability against HF needs to be assured LiPF6 −→ LiF + PF5 PF5 + H2 O −→ POF3 + 2HF

(1) (2)

C. Functional Requirements On the one hand, to ensure the electrical functionallity of both cell and sensor and, hence, to avoid short circuits between electrodes and sensor as well as to guarantee the chemical inertness, the sensor should be electrically insulated by a passivation layer. On the other hand, thermocouples can only detect relative temperature differences between the measurement point (hot junction) and the connector (cold junction)

Fig. 1. Different possibilities for sensor integration in a stack of electrodes. (a) Cell layers with the thin-film thermocouple positioned between anode current collectors. (b) Cell layers with the thin-film thermocouple positioned between anode and cathode. (c) Cell layers with the thin-film thermocouple positioned between two anodes after removing the middle cathode.

(see [12], see Fig. 2). Therefore the temperature outside the cell needs to be constantly measured with a calibrated reference sensor during operation. Therewith, only the following two options for the integration are suitable. The first solution is to directly insert the sensor between one electrode and the separator as shown in Fig. 1(b),

MARTINY et al.: DEVELOPMENT OF AN ALL KAPTON-BASED THIN-FILM THERMOCOUPLE MATRIX

Fig. 2. Schematic setup of the prototype sensor matrix integrated in the 2 Ah pouch cell with point 1 being reffered to as the left measurement point and point 2 as the right consequently.

leading to a reduced ionic flow in this cell layer. It is assumed that the capacity loss due to covering the electrode area is in the same percental range as the fraction of the covered area. This means, if the sensor area covers 30 % of the area in one layer in a cell with a total of 15 layers, the capacity is reduced by roughly 1.0 %. While the capacity loss is relatively small, peak currents and therewith hot spots that might lead to accelerated aging or safety issues could occur at the rim of the sensor. These influences are yet to be investigated. Another possibility is to remove the cathode of one of the middle layers and insert the sensor between two anodes and two separators respectively. A schematic drawing of this approach is shown in Fig. 1(c). The advantage for this method compared to the placement between anode and cathode is that no ions flow in this layer and therewith possible peak currents on the rim of the sensor can be eliminated. The main disadvantage of this approach is a higher loss in capacity as a whole active layer will be sacrificed. The capacity loss for a cell with originally 15 layers of electrodes is in the range of 6.7 %. Aside from the higher capacity loss, another disadvantage is the possible damage to other layers when removing the cathode from the assembled cell stack. For this reason the solution as shown in Fig. 1(b) is chosen for the here presented proof of concept. D. Sensor Positioning The position of the sensor needs to be chosen such that the ionic flow between the active materials of anode and cathode is affected as little as possible. Therefore, placing the sensor between a single coated, double layered current collector as shown in Fig. 1(a) is considered to be the best solution. However, as commercial cells mostly contain double coated electrodes for an increased energy density, this option cannot be realized in the approach described here. E. Resulting Design After taking all requirements mentioned above into account, a Kapton™ 100 MT film from Dupont with a thickness of 25 μm was chosen as a substrate for the sensor. Thanks to its relatively high thermal conductivity amongst polymers of 0.37 W/mK, the influence on the effective thermal resistance of the cell can be minimized. With a glass transition temperature of almost 400 ◦C and no melting point, it furthermore provides an excellent stability and consequently the ability to track the temperature even under abusiv scenarios.

3379

As mentioned above, the used sensor type is a thermocouple matrix with different measurement points and a single common potential. Fig. 3 shows a prototype of this sensor with two measurement points. Looking into the thermoelectric potential of commonly used metals, also known as the Seebeck coefficient, as studied in [13], copper (Cu) and nickel (Ni) provide a good voltage resolution with 22.5 μV/K compared to other metal pairs. Although the thin-film Seebeck coefficient is usually smaller than the bulk coefficient according to [14] and [15], the calculated thin-film coefficient according to (3) for a film thickness of 120 nm with 22.36 μV/K is close to the bulk coefficient and discrepancies are therewith considered to be negligible. 3 λ U (1 − p) (3) SF = S B 1 − 8 t 1+U In this equation, SF and S B are the Seebeck coefficients of the thinfilm and of the bulk material, p is the scattering coefficient of the film surface, t the thickness of the film and λ the electron mean free path. According to the Bloch quantum theory of electrical conduction in metals, it can be assumed that U = (∂ ln λ)/(∂ ln E) E=ζ = 2, with E as the energy function and the fermi energy ζ . The scattering coefficient can be assumed to be p = 0 [14]. With these assumptions (3) can be simplified to (4). 1 λ (4) SF = S B 1 − 4 t The electron mean free path λ is approximated with equation (5) for solids by [16]. λ=

1 n ion πr 2

(5)

with n ion as the number density of ions and r as the ion radius. The ion radius depends on the measuring method. Therefore, the values given by [13] and used for the calculation of the thin-film Seebeck coefficient are only estimated reference values. The ion radius is taken for the fully charged state. n ion can be calculated according to equation (6) with the mass density ρm , Avogadro’s number N A and the molar mass M. [16]. n ion =

ρm ∗ N A M

(6)

III. S ENSOR P RODUCTION For the deposition of the metal layers magnetron sputtering in argon (Ar) atmosphere is applied. Before the deposition the base pressure of the chamber is pumped to below 5 μPa and the chamber is heated overnight to eliminate residual water molecules in the system and outgassing of the substrate. After the Kapton MT substrate is clean sputtered, first Ni and then Cu are sputtered to the substrate through an aluminum shadow mask. The sequence is thereby crucial, as Cu is much more sensitive to oxidation and a copperoxide layer between both metals would influence the electrical contact as well as the Seebeck coefficient between both layers. The deposition of the Ni-Cu bilayers is conducted at a constant

3380

Fig. 3. Sensor matrix prototype with a substrate area of 2 cm × 6 cm, a common Cu potential in the middle and two measurement points (circled) in the calibration setup.

Ar flow rate of 60 sscm and a pressure of 0.4 Pa. The power density in both, Cu and Ni, targets is set to 4.4W/cm2 . The total thickness of the metal layers is controlled at 120 nm by using a surface profiler. The authors found a significant decrease in adhesion and an increasing susceptibility of the metal layers, especially of Cu, on the Kapton substrate with an increasing layer thickness. Therefore, the thickness of 120 nm was chosen in order to optimize the adhesion between the metal layers and the Kapton MT substrate, as well as the flexibility and reliability of the thin-film sensor. It provided good and stable results. For the resulting sensor matrix Cu provides the common potential. Fig. 3 shows a first prototype with two measurement points (circled) and a common copper potential. The total size of the sensor substrate is 2.5×6.5 cm. The width of the sputtered metal structure is 2 mm and the crosssectional points that are circled in red in Fig. 3 have a diameter of 4 mm. The used shadow mask and production process were thereby identical to those used for the parylene coated sensor described in [17]. Only the film thickness of the metal layers was reduced from originally 200 nm to 120 nm in order to optimize the adhesion on the substrate. Influences on the Seebeck coefficient as described in Section II-E are assumed to be negligible according to (3). For the coating of the sensor, two different approaches have been investigated. The first approach uses Parylene C from Specialty Coating Systems as a protective layer and has been discussed in the earlier version of this paper (see [17]). While coating with parylene has the advantage of providing a very thin layer in the micrometer range, it showed drawbacks for the integration in cells. The parylene layer suffered damage during the integration of the sensor in the cell which was investigated by a post-mortem analysis of the cell. Moreover, the measured voltage at the cold junction of the integrated sensor had increased by a factor of 320 during operation compared to the initial callibration voltage. However, the measured voltage over temperature of the integrated sensor could not be reproduced. Consequently, the sensor in the cell did not provide reliable temperature data. In further steps a different approach has been chosen therefore.


In the approach described here the sensor is covered with the same Kapton 100 MT that is used as a substrate for the thin-film thermocouple. Due to the increased thickness of the sensor of 54 μm compared to the one with the parylene coating (27 μm), the overall thermal conductance between the sensor and surrounding is lower. The thermal stability of the whole sensor increases significantly up to almost 400 ◦C thanks to the all Kapton-based design, whereas parylene has a melting point of 290 ◦C. Additionally, a more homogeneous setup on both sides of the sensor decreases the possibility of side-effects, such as unsymmetric temperature detection, that could emerge due to different thermal conductances on each side of the sensor. The Kapton cover is attached to the Kapton substrate with a Kapton tape. In order to maintain the thermal conductance between sensor and surrounding, it is important not to cover the measurement points with the tape so that they are exposed to the surrounding temperature as good as possible. IV. E XPERIMENTS AND R ESULTS For proving the concept and for investigating the influence of the sensor on the cell behavior and vice versa, different experiments are conducted. For the in-cell measurements in the earlier version of this paper, self-made pouch bag cells with two layers of LiCoO2 (LCO) cathodes and graphite anodes were used [17]. In contrast to that, commercial SPB605060 pouch cells from Enertech are used for the here described purpose and to further eliminate the influence of discrepancies due to poor cell manufacturing processes on the laboratory scale. These cells also contain LCO cathodes with aluminum (Al) current collectors and graphite anodes with Cu current collectors and a LiPF6-based organic carbonate electrolyte and are therewith similar to the previousely self-made cells. With a nominal capacity of 2 Ah and 15 layers of electrode pairs they are much larger though. The area of the electrodes is 4.5 cm × 5.1 cm for the cathode and 4.6 cm × 5.2 cm for the anode respectively. For an initial assessment of the occuring temperature spread within this cell, Figures 4(a) to 4(c) display the simulated temperature distribution at 100 % depth of discharge (DOD) according to cycling data from a 1C and 2C discharge. 1C thereby represents the charge or discharge current equal to the cell capacity in one hour. If the here mentioned 2 Ah cell is discharged with 2C, this equals a discharge current of 4 A and a discharge time of 30 min from 100 % to 0 % state of charge (SOC). The simulation studies are based on a linear interdependency of cell polarization and exchange current density as pointed out in [18] and [19] in order to estimate the temperature distribution within the electrodes [20] and, hence, the entire cell geometry [21]. The studies are carried out with the aid of COMSOL Multiphysics 4.3b. The presented simulation results show that even for a comparably small cell under normal operation conditions, inhomogenities in the temperature distribution will occur. While the temperature gradients are clearly not as significant as in large size pouch cells, this cell is still very well suitable for a proof of concept.


3381

Fig. 5. Temperature at a PTC heating element measured with a K-type thermocouple in comparison to the calculated temperature based on the linear Seebeck coefficient at the right and left measurement point.

contacts to a data logger from Agilent. A positive temperature coefficient (PTC) heating element is placed on the measurement points and heated up to 110 ◦C. One reference K-type thermocouple is attached to the heating element with a sticky tape whereas a second one is attached to the cold junction to measure the environmental temperature. The voltage of the sensor and the two temperatures of the reference thermocouples are recorded at the data logger with a sampling rate of 2 Hz. The resulting Seebeck coefficient for the used Cu/Ni films was thereof calculated according to (7) [12], where α corresponds to the thinfilm Seebeck coefficient SF from (4). V is the voltage difference between hot and cold junction and T the temperature difference between the two junctions. α was found to be 8.9 μV/K for the left measurement point (number 1 in Fig. 2) and 7.79 μV/K for the right point (number 2 in Fig. 2) respectively, and is therewith smaller than the estimated thin-film coefficient calculated in Section II-E. V = αT

Fig. 4. Simulated temperature distribution in the SPB605060 pouch cell at a 100% DOD at a constant discharge rate of 2C. (a) Simulated temperature distribution within the entire cell stack. (b) Simulated cross-sectional temperature distribution within the cell stack along the tabs and within the symmetry plane (yz-plane). (c) Simulated temperature distribution in the middle plane (xy-plane) of the cell stack where the sensor matrix is going to be integrated.

A. Sensor Calibration The calibration of the Kapton covered sensor is performed before it is integrated into the lithium ion pouch cell. The setup for calibration can be seen in Fig. 3. The cold junction (see Fig. 2) of the sensor is thereby contacted via spring

(7)

The origin of the discrepency between calculated and measured values cannot entirely be determined. 3D effects due to the surface roughness of the Kapton substrate are porbably one influence that leads to a non-optimal electrical and thermal conduction between the two metal layers. Another reason might be that the calculations in (4) are based on various assumptions and may not be fully accurate. The discrepancy between the two measurement points is probably caused by variations or impurities in the production process which are inevitable on the laboratory scale. For the required temperature range, the linear approximation is in good accordance with the measured temperature data. Fig. 5 shows the temperature at both measurement points detected with a K-type thermocouple at the heating element in comparison to the temperature calculated from the voltage values at these measurement points. For the calculation the linear Seebeck coefficients according to (7) were applied. During the heating phase this linear approximation shows slightly bigger discrepancies to the measured data compared to the consecutive cooling phase. This can be seen in the cutout of Fig. 5. The divergency can probably be attributed to the quick heating of the heating element compared to the relatively slow response time of the used data logger.

3382

Fig. 6. Voltage over temperature for two measurement points with common Cu potential in the uncovered stage and with the ready-for-integration Kapton cover.


Fig. 7. Temperature over time for the left heated and the right unheated measurement point from a thermocouple matrix with common Cu potential.

This effect will be investigated in further research activities. Still, an average discrepancy between calculated and measured values of 0.170 K with a standard deviation of 1.488 K for the left measurement point and an average discrepancy of 0.072 K with a standard deviation of 1.408K for the right measurement point is in a good range. This also underlines the improvement of the here described development compared to the parylene coated sensor where only an average discrepancy of 0.44 K with a standard deviation of 1.58 K could be realized. B. Influence of Common Potential and Coating on Temperature Measurement In Fig. 6 voltage as a function of temperature is displayed for two measurement points with a common Cu potential and a Kapton cover in comparison to the non-covered measurement points. It shows that due to the cover the Seebeck coefficient is slightly smaller and that the spread between the two measurement points increases. This is probably due to the unevenness of the attached Kapton tape for thightening the sensor with its Kapton cover. However, when looking into the linearity of the voltage-over-temperature ratio, it has significantly increased compared to the first prototype with parylene coating that has been described in [17]. The norm of residuals for the linear approximation of the Kapton covered sensor has been improved by 88 % in comparison to the parylene covered sensor from 0.00093 to 0.00011 for the left measurement point and by 51 % to 0.00045 for the right measurement point respectively. To see the influence of the common potential on each measurement point, further tests have been conducted, where only one measurement point has been heated while the other point was left at room temperature. Fig. 7 shows temperture data from this measurement. The slight increase in temperature in the beginning at the cold measurement point can be referred to unpreventable heat transfer mechanisms from the heating element to the cold measurement point. Reciprocal effects due to the common potential could not be observed.

Fig. 8. Comparison of Voltage over SOC for coin cells with a half Kapton disk between the electrodes and in normal setup, cycled at 0.25C.

To ensure this, the sensor substrate together with the Kapton tape has been immersed in electrolyte for one week and the weight, thickness and density of the sensor has been recorded before and after that. No changes in these parameters could be observed. D. Sensor Positioning Test To investigate the influence of a sensor between electrodes on the cell capacity, three different types of 2016 coin cells are assembled with LCO cathodes, graphite anodes and a Celgard 2325 separator. The chosen electrode diameter is 16 mm and the separator diameter 18 mm. 100 μL of LiPF6 -based liquid electrolyte are employed. One type of coin cells is setup in the usual way with only electrodes and separator, while a second type incorporates a Kapton MT disk with the same diameter as the separator which is integrated between the electrodes. A third type is built with only half a disk of Kapton between the electrodes, covering therewith 50 % of the electrode area. While the cells with the full Kapton disk do not function due to the low porosity of the Kapton, the other two types are cycled with 0.25C for several cycles. By displaying the voltage as a function if SOC, Fig. 8 shows that the influence of the inserted Kapton on the cell behavior is relatively low. Hence, the integration of Kapton is regarded to be acceptable for the herein stated purpose.

C. Stability Test

E. Integration Test

As mentioned above, impermeability of the sensor materials is essential to avoid any interference between cell and sensor.

In a second step, the coated sensor is integrated in the commercial 2 Ah-cell as described above to investigate the


Fig. 9. Cell with integrated sealed-in sensor tightened with Kapton tape on the left and the regular cell tabs on the right.

influence of the sensor on the cell and vise versa. The cell is therefore opened within a glovebox under Ar atmosphere by cutting the pouch bag open on the opposite side of the tabs. The sensor is then inserted for 3.5 cm into the cell between the middle pair of electrodes as shown in Fig. 2, covering therewith an area of 3.5 × 2.5 cm or 43.6 % of one electrode pair, which is roughly 1.5 % of the entire electrode area. It is therewith expected that the total capacity of the cell will decrease in the same percentage range. The measurement points as shown in Fig. 2 are inside the cell while the cold junction remains on the outside. The whole cell pack is then sealed into another pouch bag with the areas near the sensor interconnections additionally reinforced with Kapton tape. Fig. 9 shows the cell prepared in such manner. After assembly, the cell is cycled at low currents of 0.5C for charging and 0.2C for discharging respectively to proof the functionality. Fig. 10 shows the voltage as a function of SOC for the described cycle of a cell with an integrated sensor in comparison to a cell without sensor. The behavior of both cells is in very good agreement. The capacity of the cell with the integrated sensor is 3.9 % lower than the one of the cell without a sensor and therewith larger than expected from the electrode covering. This is probably caused by the reduced physical contact between the electrodes due to the integrated sensor. It shows that the electrode layer becomes probably inactive as one electrode layer equals roughly 3.3 % of the overall cell capacity. Further losses may be attributed to electrolyte loss during assembly. Additional overpotentials and/or polarization effects emerge as a consequence of the sensor integration. This is indicated by a higher voltage profile during charge and lower voltage profile during discharge to reach the same voltage thresholds of 4.2 V as the upper cut-off limit and 3.0 V as the lower limit respectively, as displayed in Fig. 10. This further reduces the energy efficiency of the battery cell by requiring more electrical work to charge the cell, while the amount of electrical work that the cell can yield during discharge is reduced. Both cells have been cycled for 15 cycles with the above mentioned cycling rates. The capacity showed an overall stable characteristic in both cases and a further effect of the sensor on the cell could not be observed.

3383

Fig. 10. Voltage over SOC of a cell with integrated sensor in comparison to a cell without sensor for one cycle with 0.5C charge and 0.2C discharge rate.

After five cycles the same experiment as performed for calibration purposes was repeated, but with the sensor still integrated in the cell. A significant increase in the voltageover-temperature ratio could be observed, similar to the parylene coated sensor. While the measured voltage on the sensor during calibration reached 600 μV at 95 ◦C, the measured voltage after cycling reached a value as high as 120 mV at 40 ◦C. This also resulted in unreproducible voltage over temperature data. As this behavior is similar to the behavior of the parylene coated sensor, which has been disassembled from the cell, it is assumed that a leakage in the insulating cover of the sensor is the reason. This shows again the high importance and difficulty of a persistent and stable coating of the sensor so that electrolyte leakage can be avoided. In the next step, the authors plan to build the sensor with liquid polyimide on both sides of the sensing area to ensure the impermeability of the sensor. V. C ONCLUSION The authors presented a novel approach for designing a thinfilm thermocouple matrix. Copper and nickel as thermocouple metals showed a good voltage-over-temperature dependence, even though the measured voltage was smaller than the calculated thin-film coefficient. The application of copper as a common potential for different measurement points provided very good results and during tests with different temperatures at each point a reciprocal effect between the measurement points could not be observed. The coating of the sensor with Kapton showed an increase in linearity of the voltage-over-temperature dependence during calibration and a good stability during different measurements compared to the before presented approach with parylene. Due to the higher thickness of the Kapton cover, the Seebeck coefficient decreased. Measurements showed still reliable and reproducible results. The sensor has been integrated in a commercial 2 Ah pouch cell and the cell has been cycled for 15 cycles. While no influence from the sensor on the cell could be observed, after some cycles the adhesion of the sensor cover weakened which resulted in a leakage of electrolyte. The experiments showed the necessity of an improved manufacturing process to ensure long-term sensing stability. In a next step, the sensor will be

3384


built from liquid polyimide giving the possibility to integrate the metal layers directly in the substrate. ACKNOWLEDGMENT The authors would like to thank A. Lambrecht from DuPont E&C - Circuit & Packaging Materials for providing us the Kapton™ substrate. Thanks also to L. Moraleja who constantly helps in setting up the experiments. R EFERENCES [1] N. Martiny, P. Osswald, C. Huber, and A. Jossen, “Safety management for electric vehicle batteries in a tropic environment,” in Proc. Electr. Veh. Symp., Los Angeles, CA, USA, May 2012. [2] M. S. K. Mutyala, J. Zhao, J. Li, H. Pan, C. Yuan, and X. Li, “Insitu temperature measurement in lithium ion battery by transferable flexible thin film thermocouples,” J. Power Sources, vol. 260, pp. 43–49, Aug. 2014. [3] Z. Li et al., “Examining temporal and spatial variations of internal temperature in large-format laminated battery with embedded thermocouples,” J. Power Sour., vol. 241, pp. 536–553, Apr. 2013. [4] C.-Y. Lee, S.-J. Lee, M.-S. Tang, and P.-C. Chen, “In situ monitoring of temperature inside lithium-ion batteries by flexible micro temperature sensors,” Sensors, vol. 11, pp. 9942–9950, Oct. 2011. [5] X. Feng et al., “Thermal runaway features of large format prismatic lithium ion battery using extended volume accelerating rate calorimetry,” J. Power Sources, vol. 255, pp. 294–301, Jun. 2014. [6] A. Rheinfeld, S. Erhard, S. Kosch, and A. Jossen, “Modelling of temperature distribution within a large format Li-ion battery during discharge and its implications towards internal temperature sensor placement for designing safe battery systems,” in Proc. Kraftwerk Batterie, Münster, Germany, 2014. [7] A. Reinfelder, N. Martiny, and A. Jossen, “Thermal in-cell measurement for li-ion pouch cells,” in Proc. Conf. Future Autom. Technol., Mar. 2012. [8] C.-Y. Lee, W.-J. Hsieh, and G.-W. Wu, “Embedded flexible microsensors in MEA for measuring temperature and humidity in a micro-fuel cell,” J. Power Sources, vol. 181, no. 2, pp. 237–243, Jul. 2008. [9] S. T. Ali, J. Lebaek, L. P. Nielsen, C. Mathiasen, P. Moller, and S. K. Kaer, “Thin film thermocouples for in situ membrane electrode assembly temperature measurements in a polybenzimidazole-based high temperature proton exchange membrane unit cell,” J. Power Sources, vol. 195, no. 15, pp. 4835–4841, Aug. 2010. [10] S. He, M. M. Mench, and S. Tadigadapa, “Thin film temperature sensor for real-time measurement of electrolyte temperature in a polymer electrolyte fuel cell,” Sens. Actuactors A, Phys., vol. 125, no. 2, pp. 170–177, Jan. 2006. [11] S. F. Lux, I. T. Lucas, E. Pollak, S. Passerini, M. Winter, and R. Kostecki, “The mechanism of HF formation in LiPF6 based organic carbonate electrolytes,” Electrochem. Commun., vol. 14, no. 1, pp. 47–50, 2012. [12] J. Fraden, Handbook of Modern Sensors: Physics, Design and Applications, 3rd ed. New York, NY, USA: Springer, 2004. [13] H. Stöcker, Ed., Taschenbuch der Physik, 6th ed. Frankfurt, Germany: Deutsch Harri GmbH, 2010. [14] K. L. Chopra, Thin Film Phenomena. New York, NY, USA: McGraw-Hill, 1969. [15] X. Zhang, H. Choi, A. Datta, and X. Li, “Design, fabrication and characterization of metal embedded thin film thermocouples with various film thicknesses and junction sizes,” J. Micromech. Microeng., vol. 16, no. 5, pp. 900–905, 2006. [16] P. A. Tipler and G. Mosca, Physics for Scientists and Engineers, vol. 6. San Francisco, CA, USA: Freeman, 2008, [17] N. Martiny, J. Geder, Y. Wang, W. Kraus, and A. Jossen, “Development of a thin-film thermocouple matrix for in-situ temperature measurement in a lithium ion pouch cell,” in Proc. IEEE Sensor, Nov. 2013, pp. 1–4. [18] W. Tiedemann and J. Newman, “Current and potential distribution in lead-acid battery plates,” in Proc. Symp. Battery Des. Optim., 1979. [19] K. Kwon, C. Shin, T. Kang, and C.-S. Kim, “A two-dimensional modeling of a lithium-polymer battery,” J. Power Sources, vol. 163, no. 1, pp. 151–157, 2006.

[20] U. Kim, C. B. Shin, and C.-S. Kim, “Effect of electrode configuration on the thermal behavior of a lithium-polymer battery,” J. Power Sources, vol. 180, no. 2, pp. 909–916, 2008. [21] J. Yi, U. Kim, C. Shin, T. Han, and S. Park, “Three-dimensional thermal modeling of a lithium-ion battery considering the combined effects of the electrical and thermal contact resistances between current collecting tab and lead wire,” J. Electrochem. Soc., vol. 160, no. 3, pp. A437–A443, 2013.

Nora Martiny received the Degree in electrical engineering and information technology from Technische Universität München (TUM), Munich, Germany, in 2010. From 2009 to 2010, she worked parallel to her studies in the research project Snookie within the robotics cluster Cotesys at TUM and, from 2009 to 2010, at the Robotics Research Laboratory, Nanyang Technological University, Singapore. Her research within TUM CREATE focuses on the development of safety monitoring techniques for lithium-ion pouch cells.

Alexander Rheinfeld received the Degree in mechanical engineering from Technische Universität München (TUM), Munich, Germany, in 2012. He conducted his graduation thesis in collaboration with Audi in the field of thermal management of battery electric vehicles. In 2012, he joined the Institute for Electrical Energy Storage Technology at TUM. He is working on the modeling of safety behavior aspects of lithium-ion cells and is currently contributing to the project SafeBatt.

Jan Geder received the degree in chemical engineering from the University of Ljubljana, Ljubljana, Slovenia, in 2010. After gaining experience in process technology, he is with Siemens AG, Munich, Germany, and Forschungszentrum Jülich, Jülich, Germany, and is currently pursuing the Ph.D. degree at TUM CREATE. He is currently working on lithium-ion battery safety research. The topics of his research range from safety properties and performance of lithium-ion battery materials to the development of new characterization methodologies.

Yuxi Wang received the degree in metrology from Tianjin University, Tianjin, China, in 2007, and the Ph.D. degree in material engineering from Nanyang Technological University, Singapore, in 2013. His research interests include thin-film fabrication and characterization for various industrial applications. In 2012, he joined TUM CREATE Ltd., Singapore, as a Research Fellow, where he was involved in refuelable lithium–oxygen batteries for nextgeneration electrical vehicle applications.

Werner Kraus is a Researcher with Technische Universität München (TUM), Munich, Germany. He received the Diploma degree in electrical engineering and the Ph.D. degree from TUM in 1997 and 2008, respectively. Though he studied analog and mixed-signal circuits for his Ph.D. thesis, he has been devoted to investigations on semiconductor technology processes. His major interests in thin-film technology are strongly correlated to sensors and their applications.

Andreas Jossen was a Research Fellow with the Chair of Theory of Electrical Engineering after studying electrical engineering from the University of Stuttgart, Stuttgart, Germany. He has worked in the field of energy storage for photovoltaic systems. Prior to becoming a Full Professor at Technische Universität München, Munich, Germany, he was with the Centre for Solar Energy and Hydrogen, Ulm, Germany, until 201, heading up a workgroup on battery system technology. His chair explores electrochemical energy storage for various applications.