Journal of The Electrochemical Society, 149 共3兲 A302-A306 共2002兲

A302

0013-4651/2002/149共3兲/A302/5/$7.00 © The Electrochemical Society, Inc.

A Nonaqueous Asymmetric Hybrid Li4 Ti5 O12 ÕPoly„fluorophenylthiophene… Energy Storage Device Aurelien Du Pasquier,a,*,z Alexis Laforgue,b,** Patrice Simon,b,** Glenn G. Amatucci,a,* and Jean-Francois Fauvarqueb,* a

Telcordia Technologies, Red Bank, New Jersey 07701, USA CNAM, Laboratoire d‘Electrochimie Industrielle, 75003 Paris, France

b

An asymmetric hybrid energy storage device has been built using a lithium titanate intercalation anode and the pseudocapacitive electronically conducting polymer poly共fluorophenythiophene兲. Results show that the device operates in the power range of electrochemical double-layer capacitors, while having three to four times their energy density, and a voltage profile and selfdischarge characteristics typical of a battery. The cycle life is also intermediate between that of a battery and a supercapacitor, in the range of several thousands of cycles. © 2002 Telcordia Technologies. 关DOI: 10.1149/1.1446081兴 All rights reserved. Manuscript submitted June 25, 2001; revised manuscript received October 5, 2001. Available electronically January 30, 2002.

Electrochemical double-layer capacitors 共EDLCs兲, or supercapacitors, are electrochemical devices that store energy as charge at the surface of the electrodes, forming the so-called double-layer capacitance. There is no electron transfer associated with the charge movement, and no chemical modification of the active materials. This is defined as capacitive electrochemical energy storage. Therefore, the kinetics of charge storage and restitution are very fast and very reversible. This results in high power capability in the range of 1 to 10 kW/kg, and extended cycle life, in the range of 100 to 1000 thousands of cycles. The main drawbacks of supercapacitors are their fast self-discharge, and their low energy density and poor voltage profile, where a large part of the energy is recovered below 1 V, a feature not compatible with most applications. Typical electrode active materials for supercapacitors are activated carbons of high Brunauer, Emmett, Teller 共BET兲 surface area. On the other hand, batteries are energy storage devices where electrochemical reactions occur in the bulk of the electrodes, modifying their degree of oxidation, chemical composition, and lattice parameters. This is defined as faradaic electrochemical energy storage. This results in higher energy densities, lower power due to the slower kinetics of the electrochemical reactions and the existence of a charge transfer resistance, and lower cycle life due to the structural changes caused by electrochemical reactions. A typical cycle life for batteries is in the range of 100 to 1000 cycles. There is a class of active materials which display characteristics that are intermediate between those used in supercapacitors and batteries. Those are mostly materials which undergo surface reactions associated with electron transfer and double-layer formation, a behavior defined as pseudocapacitive. Typical examples of pseudocapacitive materials are noble oxides, such as RuO2 , and electronically conducting polymers, such as polythiophenes and polypyrrole. In the case of electronically conducting polymers, the pseudocapacitive behavior originates from the fact that p- and n-doping are faradaic processes 共oxidation or reduction of the polymer backbone兲, associated with adsorption or desorption of counterions from the solution in order to compensate the charge modification on the polymer. Pseudocapacitive materials are attractive because they combine fast kinetics 共e.g., high power兲, and specific capacitance values greater than the double-layer capacitance purely associated to the surface BET area. These materials provide an opportunity to create devices that bridge the gap between supercapacitors and batteries. It is possible to build an electrochemical capacitor that combines a pseudocapacitive electrode with a capacitive counter electrode. For

* Electrochemical Society Active Member. ** Electrochemical Society Student Member. z

E-mail: [email protected]

instance, U.S. patent 6,222,723 describes the use of activated carbon 共capacitive兲 electrode and noble oxide 共pseudocapacitive兲 counter electrode in aqueous electrolyte. This work presents results of a different type of hybrid energy storage device utilizing a lithium intercalation 共faradaic, nonpseudocapacitive兲 anode and a pseudocapacitive cathode in a nonaqueous electrolyte. The asymmetric hybrid concept.—Telcordia Technologies has recently presented the concept of a nonaqueous asymmetric hybrid energy storage device utilizing a lithium intercalation anode and an activated carbon cathode.1,2 Nanosized L4 Ti5 O12 共hereafter, LTO兲 has been found to meet all the requirements for such a hybrid device: extended cycle life, good capacity, low voltage with respect to standard hydrogen electrode 共⬍⫺1 V vs. SHE兲, good stability during fast lithiation 共charge cycle in the hybrid cell兲, and excellent rate capability to complement the positive electrode. The use of Li4 Ti5 O12 lithium intercalation anode enables the shift of the average voltage of the device from 1.5 V in a conventional EDLC to 2.25 V, with the end of discharge occurring around 1.5 V, as opposed to 0 V in an EDLC. This results in a significant increase in energy density. 20 Wh/kg has been demonstrated in an LTO/C device, with 90% capacity utilization at 10C charge/discharge rates.3 Oxidation of the electrolyte at the surface of the activated carbon has been pointed out as the main limitation of the device, causing its self-discharge in as similar fashion as EDLCs. The use of a pseudocapacitive organoredox cathode material of smaller BET specific surface area is projected to considerably improve the situation. Furthermore, since the charge storage mechanism is not purely capacitive, but also caused by the p-doping of the cathode, better charge retention is expected to occur upon storage. A second reason for replacing the activated carbon by a pseudocapacitive cathode is to increase the energy density and power of the device, due to the higher specific capacity and average voltage of those materials as opposed to activated carbons. Positive electrode: poly(fluorophenylthiophene), PFPT.—PFPT was originally proposed by Rudge et al.4 for pseudocapacitive/ pseudocapacitive devices using both the n doping of PFPT in the negative electrode, and the p doping in the positive. However, the work of Mastragostino et al.5 pointed out the limitations in cycle life and energy density of PFPT used in the n-doped state. Good cycling performance of the p doping in PFPT has recently been shown by Laforgue et al.6 in an EDLC/pseudocapacitive device where activated carbon was used as the negative electrode and PFPT as the positive electrode. To our knowledge, the electrochemical behavior of PFPT has never been examined in a lithium-based electrolyte. The use of PFPT to replace activated carbon for the positive electrode of a supercapacitor has several advantages. The pseudocapaci-

Journal of The Electrochemical Society, 149 共3兲 A302-A306 共2002兲

A303

Figure 1. Chemical synthesis procedure of the PFPT.

tive reaction associated with the p doping of the polymer generates a higher average output voltage than the ion adsorption in a double layer of an activated carbon. This results in an increase in energy density, and also in the energy quality, because most of the power is recovered at higher voltages than in a carbon-carbon capacitor. Also, the BET area is two orders of magnitude lower than in an activated carbon. This is an advantage for reducing gassing and self-discharge of the devices. Finally, the specific capacitance is usually two to three times that of activated carbons. These advantages made PFPT a good choice to replace the positive electrode activated carbon in an asymmetric hybrid device. Experimental Preparation of PFPT.—The monomer was synthesized by a coupling reaction of 3-bromothiophene and 4-fluorophenylmagnesium bromide in tetrahydrofuran 共THF兲 with NiCl2 共diphenylphosphinopropane兲 as catalyst. This preparation method is well known for the synthesis of 3-alkylthiophenes.7-9 The yield for this reaction was more than 90% in weight. The monomer was then chemically polymerized to obtain a powder that can be industrially processed into composite electrodes. The reaction proceeded through a direct oxidation using four equivalents of anhydrous FeCl3 as oxidant 共Fig. 1兲. As FeCl3 is very reactive, the reaction was performed in anhydrous chloroform under argon atmosphere. The polymerization was carried out during 20 h and stopped by adding methanol into the reaction medium. The powder was then washed several times with methanol to dedope the polymer and get rid of the oligomers. Its BET surface area was 24.5 m2 /g, measured by nitrogen adsorption with a Micromeritics ASAP2000 apparatus. This method of polymerization has already been used to prepare thiophene derivatives.10-13 It leads to high molecular weight polymers14-17 and around 80% of regio-regular 共head-to-tail兲 couplings.18,19 Negative electrode.—As shown by Ohzuku et al., LTO is one of the few lithium intercalation compounds in existence that undergoes little appreciable expansion or contraction during the lithium insertion/reinsertion process.20 This results in an outstanding cycle life, as shown on Fig. 2 for a material prepared at 800°C and cycled at a rate of C/2. The two-phase reaction results in a flat intercalation plateau at 1.5 V vs. (Li⫹, Li), and the theoretical capacity is 175 mAh/g. The intercalation voltage is sufficiently low to provide a good energy density when used with a pseudocapacitive cathode having a potential of 3-4.5 V vs. (Li⫹, Li), and provides a safety window against Li plating during fast lithiation. A nanomaterial Li4 Ti5 O12 was specially designed for this application, and demonstrated superior rate capability during lithiation. Figure 3 compares the lithiation rate capability of the nanosized LTO to conventional crystalline LTO prepared by solid-state synthesis. Electrochemical measurements.—Cyclic voltametry experiments on PFPT were performed using a PARC EG&G model 273A potentiostat controlled by a computer using the software M270. Cyclic voltametry measurements were performed in a three-electrodes cell, with a platinum foil as counter electrode and a Ag⫹/Ag electrode as reference electrode 共consisting in a Ag wire in a solution of 10⫺2 M Ag⫹; NO⫺ 3 in acetonitrile兲. All the potentials were reported against Li⫹/Li. Galvanostatic characterization of the polymer was carried out with a Biologic Technologies VMP 共versatile multipotentiostat兲. Electrochemical testing of LTO and PFPT against Li were performed on a Maccor galvanostat. All cell fabrication was performed

Figure 2. Galvanostatic cycling of a Li4 Ti5 O12 composite electrode at 1 mA/cm2 ; electrolyte 1 M LiPF6 in 共1 EC-1 DMC兲 共v:v兲; Li counter electrode.

in a ⫺80°C dew point, He-filled glove box. Stainless steel coin cells 共NRC兲, utilizing borosilicate glass fiber mat separator, were adopted for cell characterization. All electrodes tested in a lithium-based electrolyte were prepared according to Telcordia plastic battery technology. Active materials were mixed with 5% carbon black 共MMM super P兲, poly共vinylidene fluoride-co-hexafluoropropylene兲 binder 共Atofina, Kynarflex兲, and propylene carbonate plasticizer in acetone. Amounts of binder and plasticizer were dependent on the BET surface area of the active materials, and are kept proprietary. The mixture was cast and dried at 22°C for 0.5 h. Afterward, disks were punched from the freestanding tape. The disks were then placed in ether to extract the plasticizer. All electrode material characterization performed at high rates was done in such configurations that the active weights of the electrode did not result in cell current densities in excess of 10 mA/cm2 . This was to ensure that system restraints did not influence the electrochemical characterization of the material themselves. Results and Discussion PFPT testing in acetonitrile-based electrolyte.—Electrochemical characterization was performed using cyclic voltammetry and galvanostatic cycling in an electrolyte comprising 1 M NEt4 CF3 SO3 in

Figure 3. Discharge capacity as a function of rate for a nano and crystalline Li4 Ti5 O12.

A304

Journal of The Electrochemical Society, 149 共3兲 A302-A306 共2002兲

Figure 4. Voltammogram of the p-doping process of the PFPT in a composite electrode. Scan rate 20 mV/s; electrolyte 1 M NEt4 CF3 SO3 in acetonitrile.

acetonitrile. The polymer powder was processed into composite electrodes with acetylene black as electronic conducting additive and a mixture of carboxymethylcellulose 共CMC兲 and polytetrafluoroethylene 共PTFE兲 as binder.21 The current collectors were expanded stainless steel AISI 316L 共Delker 4SS7兲. Figure 4 presents the voltammogram of the PFPT in the p-doping potential range. The doping process begins at 3.8 V/Li⫹/Li with a peak at 4.38 V vs. Li⫹/Li. The doped charge is about 40 mAh/g when the doping is stopped at 4.5 V vs. Li⫹/Li and 50 mAh/g when the doping is stopped at 4.6 V vs. Li⫹/Li. On reverse scan, the dedoping peak can be observed at 4.15 V vs. Li⫹/Li. The released charge is about 38 mAh/g when the polymer is doped to 4.5 V vs. Li⫹/Li and 39 mAh/g when the polymer is doped up to 4.6 V vs. Li⫹/Li. The comparison between the doped and dedoped capacities shows that the charges doped after 4.5 V vs. Li⫹/Li are essentially due to parasitic phenomena. Charge trapping or electrolyte oxidation22,23 decreased the coulombic efficiency (Q discharged / Q charged) of the doping process from 0.95 共doping to 4.5 V vs. Li⫹/Li兲 to 0.78 共doping up to 4.6 V vs. Li⫹/Li兲. Figure 5 shows the galvanostatic curve of a PFPT composite electrode 共bound with CMC-PTFE兲 at 1 mA/cm2 in the same electrolyte. The p-doping process begins at 3.8 V/Li and the electrode potential linearly increases up to 4.5 V vs. Li⫹/Li. The dedoping process takes place linearly from 4.5 to 3.8 V vs. Li⫹/Li where

Figure 5. Galvanostatic charge/discharge profile of a PFPT composite electrode at 1 mA/cm2 (m polym ⫽ 5.6 mg); electrolyte 1 M NEt4 CF3 SO3 in acetonitrile.

Figure 6. Galvanostatic charge/discharge profile of a PFPT composite electrode at 1 mA/cm2 ; electrolyte 1 M LiPF6 in 共1 EC-1 DMC兲 共v:v兲; Li counter electrode.

end-of-discharge can be observed. The charge capacity of the polymer is 42 mAh/g for a discharge capacity of 39 mAh/g. Pseudocapacitance of the polymer, calculated from the slope of the linear part of the discharge curve, is about 270 F/g. PFPT testing in a lithium-based electrolyte.—Since no former studies have been published on the electrochemical behavior of PFPT in a lithium-based electrolyte, we chose to test it in ethylene carbonate-dimethyl carbonate 共EC-DMC兲, LiPF6 1 M with metallic lithium as the counter electrode. Acetonitrile could not be used as the solvent due to its reactivity with metallic lithium. 1 cm2 PFPT electrodes were tested at a 1 mA/cm2 constant current density with 4.3-2.5 V cutoff voltages. The goal was mainly to verify that capacity was maintained and similar voltage profile was obtained when a lithium-based electrolyte replaced the acetonitrile-based electrolyte. Indeed, a slightly higher capacity of 45 mAh/g of PFPT was measured in this system 共Fig. 6, 7兲. However, the difference could have several origins: change of the electrolyte anion size, change in the anion solvatation, change in the electrode fabrication process, or change in the testing protocols. Therefore, given these changes of experimental conditions, it was concluded that the two results obtained in two different laboratories were in fairly good agreement. Hybrid cell performance.—Hybrid cells were assembled by hot lamination of Telcordia plastic battery technology-based electrodes

Figure 7. Galvanostatic cycling of a PFPT composite electrode at 1 mA/cm2 ; electrolyte 1 M LiPF6 in 共1 EC-1 DMC兲 共v:v兲; Li counter electrode.

Journal of The Electrochemical Society, 149 共3兲 A302-A306 共2002兲

Figure 8. Effect of matching ratio of PFPT to LTO on their gravimetric capacities measured in the hybrid device.

onto aluminum grids. The electrode-current collector assemblies were further hot laminated on a microporous polyolefin-based separator. Typical test cells were 5 ⫻ 3.2 cm2 laminates packaged in a multifoil PE-Al-Surlin material. The electrolyte used to activate the cells was CH3 CN, LiPF6 1.5 M, which has higher ionic conductivity than EC-DMC, LiPF6 1 M, while still being stable at the LTO electrode potential. Two weight matching ratios of the positive to negative active materials were tested, 2.7 and 3.9. These translate in absolute capacity ratios of 0.67 and 0.97, respectively, assuming theoretic capacities of 40 mAh/g for PFPT and 160 mAh/g for LTO. Cells were charged at constant current up to 3 V. Discharge tests with a matching ratio of 2.7 indicated that the PFPT electrode was fully utilized, with a measured low rate capacity of 43 mAh/g, but only 115 mAh/g was recovered from the LTO anode, indicating that it was in excess. With a matching ratio of 3.9, 140 mAh/g were recovered from LTO while 35 mAh/g were recovered from PFPT. The theoretical matching ratio is 4, considering 40 mAh/g of PFPT and 160 mAh/g of LTO, which results in an absolute capacity ratio of 1, with full utilization of both electrodes. In both cases, the discharge profile showed continuous voltage decay from 2.75 to 2 V 共Fig. 8兲. Power and energy data of packaged hybrid cells.—The flat-plate prototype cells with matching ratios of 2.7 and 3.9 were devices of respective weights 1.7 and 2.4 g, and respective capacities of 4.7 and 6.2 mAh, for a total thickness of ca. 0.7 mm. Their power and energy data are plotted on a Ragone plot obtained by discharging the fully charged devices at increased current densities. The best results obtained were 7.4 Wh/kg at low current, and 6.1 Wh/kg at a discharge power of 560 W/kg 共Fig. 9兲. Note, however, that the cells used for these measurements were not optimized for the packaging efficiency. Since only one plate was packaged, the weight fraction of packaging material was nonnegligible. It can be calculated that a similar device made from a higher number of plates would reach 12 Wh/kg for seven plates 共Fig. 10兲. Thus, increasing the size of the device tested and minimizing the electrolyte weight can increase energy density. Such an optimization was undertaken, and Fig. 11 shows the Ragone plots obtained with optimized capacitors. A lighter electrolyte, LiBF4 2 M in CH3 CN, was used, and cells were made of larger size, by folding a 5.08 ⫻ 24.1 cm laminate in a 5.08 ⫻ 3.17 cm device. Prototypes were made with various electrode thickness. For packaged devices, 20 and 10 Wh/kg at 1000 W/kg were measured with thick electrodes. 16 and 10 Wh/kg at 2500 W/kg were measured with thinner electrodes. These results show a dramatic im-

A305

Figure 9. Ragone plots for small LTO/PFPT hybrid devices of matching ratios 2.7 and 3.9.

Figure 10. Effect of packaging efficiency on energy density of LTO/PFPT small hybrid devices.

Figure 11. Ragone plots of optimized packaged LTO/PFPT hybrid devices vs. C/C supercapacitor.

A306

Journal of The Electrochemical Society, 149 共3兲 A302-A306 共2002兲 pacitors, as shown on Fig. 13. This confirms the statement that highsurface area activated carbon cathodes are responsible in large part for the self-discharge of any type of device making use of them. Self-discharge data are both better with a matching ratio of 3.9, which is close the theoretic matching ratio of 4, assuming 40 mAh/g of PFPT and 160 mAh/g of LTO. For this cell, the reversible capacity loss at the end of the 30 day storage period was 18%. For comparison, a similar LTO/C cell had lost 50% of its original capacity after only 10 h storage.

Conclusion

Figure 12. Cycle life of LTO/PFPT hybrid devices at matching ratios 2.7 and 3.9 共10C charge, 100C discharge兲.

provement to what can be obtained in similar electrode and packaging technology for carbon-carbon supercapacitors 共Fig. 11兲. Cycling and self-discharge data of packaged hybrid devices.—Cycling tests of the packaged devices were performed at constant current of 10C for the charge, and 100C for the discharge, between 0 and 3 V. Results after 1500 cycles indicate 14% capacity loss for r ⫽ 3.9 and 28% capacity loss for r ⫽ 2.7 共Fig. 12兲. Most of the losses occurred at the beginning of cycling. Packaged cells were fully charged to 3 V and left in open circuit while the voltage decay was recorded. A fast decay of the voltage to 2.5 V was observed during the first hours, then the voltage remained constant over a period of 30 days with the matching ratio of 3.9. Voltage slightly and regularly decreased with the matching ratio of 2.7. In both cases, the self-discharge data shows a dramatic improvement compared to LTO/carbon hybrid devices or carbon/carbon superca-

Figure 13. Self-discharge of LTO/PFPT hybrid devices with matching ratios 2.7 and 3.9 vs. C/C supercapacitor.

For the first time, an asymmetric hybrid cell has been built by combining a lithium intercalation anode and a pseudocapacitive cathode made of PFPT. The attributes of such a hybrid device are intermediate between those of a battery and a supercapacitor. The voltage profile shows a good quality energy, where the usable power is between 2.75 and 1.5 V. The energy density is higher than a supercapacitor, the specific power is higher than a battery. The device can be fully charged in 6 min, and discharged at power exceeding 2000 W/kg. Self-discharge is significantly lower than in the case of supercapacitors or hybrid devices using an activated carbon-based cathode. Cycle life is good, but not as good as carbon-carbon capacitors or LTO/C hybrid devices, due to the faradaic nature of the reactions involved in PFPT. Work is still in progress to improve it. The combination of a fast-kinetics nanosized lithium intercalation anode and a pseudocapacitve organic cathode appears like a promising choice for a next generation of high energy hybrid electrochemical storage devices. Telcordia Technologies assisted in meeting the publication costs of this article.

References 1. G. G. Amatucci, F. Badway, and A. Du Pasquier, Abstract 122, The Electrochemical Society and The Electrochemical Society of Japan Meeting Abstracts, Vol. 99-2, Honolulu, HI, Oct 17-22, 1999. 2. G. G. Amatucci, U.S. Pat. 6,252,762 共2001兲. 3. G. G. Amatucci, F. Badway, A. Du Pasquier, and Tao Zheng, J. Electrochem. Soc., 148, A930 共2001兲. 4. A. Rudge, J. P. Ferraris, and S. Gottesfeld, U.S. Pat. 5,527,640 共1996兲. 5. M. Mastragostino, C. Arbizzani, R. Paraventi, and A. Zanelli, J. Electrochem. Soc., 147, 407 共2000兲. 6. A. Laforgue, P. Simon, J. F. Fauvarque, J. F. Sarrau, and P. Lailler, J. Electrochem. Soc., 148, A1130 共2001兲. 7. K. Tamao, S. Kodama, I. Nakajima, M. Kumada, A. Minato, and K. Suzuki, Tetrahedron, 38, 3347 共1982兲. 8. T. Sone, M. Inoue, and K. Sato, Bull. Chem. Soc. Jpn., 61, 3779 共1988兲. 9. J. P. Monthe´ard, J. F. Delzant, and M. Gazard, Synth. Commun., 14, 289 共1984兲. 10. R. Sugimoto, S. Takeda, H. B. Gu, and K. Yoshino, Chem. Express, 1, 635 共1986兲. 11. N. Sato, M. Rikukawa, K. Sanui, and N. Ogata, Synth. Met., 101, 132 共1999兲. 12. T. Olinga and B. Franc¸ois, Synth. Met., 69, 297 共1995兲. ¨ sterholm, and J. Korvola, Polymer, 33, 1559 13. V. M. Niemi, P. Knuuttila, J. E. O 共1992兲. ¨ sterholm, J. Laakso, P. Nyholm, H. Isotalo, H. Stubb, O. Ingana¨s, and W. R. 14. J. E. O Salaneck, Synth. Met., 28, 435 共1989兲. 15. F. Andreani, E. Salatelli, M. Lanzi, F. Bertinelli, A. M. Fichera, and M. Gazzano, Polymer, 41, 3147 共2000兲. 16. M. Kobashi and H. Takeuchi, Synth. Met., 101, 585 共1999兲. 17. R. M. Souto Maior, K. Hinkelmann, H. Eckert, and F. Wudl, Macromolecules, 23, 268 共1990兲. 18. J. Roncali, Chem. Rev., 97, 173 共1997兲. 19. P. C. Stein, C. Botta, A. Bolognesi, and M. Catellani, Synth. Met., 69, 305 共1995兲. 20. T. Ohzuku, A. Ueda, and N. Yamamoto, J. Electrochem. Soc., 142, 1431 共1995兲. 21. A. Laforgue, P. Simon, and J. F. Fauvarque, Synth. Met., To be published. 22. J. Wang, Electrochim. Acta, 42, 2545 共1997兲. 23. E. W. Tsai, S. Basak, J. P. Ruiz, J. R. Reynolds, and K. Rajeshwar, J. Electrochem. Soc., 136, 3683 共1989兲.