Appropriate methods for evaluating the efficiency and

Electrochemistry Communications 60 (2015) 21–25

Contents lists available at ScienceDirect

Electrochemistry Communications journal homepage: www.elsevier.com/locate/elecom

Short communication

Appropriate methods for evaluating the efficiency and capacitive behavior of different types of supercapacitors A. Laheäär, P. Przygocki, Q. Abbas, F. Béguin ⁎ Institute of Chemistry and Technical Electrochemistry, Poznan University of Technology, Berdychowo 4, 60-965 Poznan, Poland

a r t i c l e

i n f o

Article history: Received 30 June 2015 Received in revised form 25 July 2015 Accepted 26 July 2015 Available online 31 July 2015 Keywords: Electrical double-layer capacitor Redox supercapacitor Coulombic efficiency Energy efficiency Capacitive behavior

a b s t r a c t The development of new brands of supercapacitors (SCs) has led to a variety of energy storage mechanisms and frequently to performance overestimation or erroneous presentation of the capacitive behavior through applying mathematical relations valid only for electrical double-layer capacitors (EDLCs). This paper addresses a realistic evaluation of capacitive performance and efficiency of SCs based on carbon electrodes. The presented examples of imprecise data processing include misleading information, such as seeming discharge capacitance increase during SC aging and coulombic efficiencies of 90% for SCs involving redox processes whilst energy efficiency is only 50%. Even in typical EDLCs, energy efficiency is 5–10% lower than coulombic efficiency. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Lately, the R&D on supercapacitors (SCs) has bloomed, as these devices are considered very promising for several applications such as hybrid and electric vehicles, power tools, regulation of electric energy from renewable sources, etc. [1]. The pursuit for high energy/high power combination has led to a variety of charge storage mechanisms in SCs. As a result, some systems do not follow precisely the SC's characteristic by definition—a constant capacitance throughout the applied potential range, and the line between classification of high-energy SC and high-power battery has become rather blurred. Therefore using mathematical relations developed for ideal electrical double-layer capacitors (EDLCs) should be carefully reconsidered. Different SC electrode active materials have been recently revisited, including suggested calculation approaches in association with the individual electrode processes [2]. The fundamental difference between pseudocapacitive and faradaic electrodes, as well as incorrect/meaningless characterization of faradaic materials by “capacitance”, was clearly re-explained in reference [3]. A critical review accentuates that the redox electrolyte interface displays capacity instead of capacitance and should not be assumed to follow EDLC relations [4]. Another important SC characteristic—efficiency—needs to be evaluated properly. Cycle efficiency is defined as the ratio of energy delivered by a capacitor to the energy supplied to it during a specified cycle. In case of an ideal EDLC, perfectly linear galvanostatic curves are recorded, and the energy efficiency (ratio of areas under the discharge/charge ⁎ Corresponding author. E-mail address: [email protected] (F. Béguin).

http://dx.doi.org/10.1016/j.elecom.2015.07.022 1388-2481/© 2015 Elsevier B.V. All rights reserved.

curves, multiplied by current) can be simply expressed by the coulombic efficiency (ratio of discharge/charge durations). However, for many SCs, especially those involving charge transfer reactions, efficiency determined by discharge/charge durations is overestimated due to nonlinearity of the cell potential profiles. In this paper, we demonstrate that efficiency calculated from the energy ratio is inferior to that estimated by the time ratio for activated carbon (AC) based redox SCs as well as EDLCs with various electrolytes. Especially, the state-of-health after aging is remarkably worse when evaluated by energy efficiency rather than coulombic efficiency. Similarly, the choice of capacitance evaluation method can strongly affect the interpretation of processes during aging of SCs. Instead of simplifying to “linear behavior”, integration of galvanostatic curves must be applied for redox electrolyte SCs, to avoid overestimating the capacitive performance and to reveal the fair energy recovery. Overall, whatever the type of capacitor storage mechanism, data recovered from integration provides the most reliable set of capacitor characteristics. 2. Experimental The carbon materials used were an activated carbon YP-80F (Kuraray; SBET = 2270 m2 g−1), a carbide-derived carbon C(Mo2C) (synthesized by chlorination of Mo2C at 800 °C [5]; SBET = 1680 m2 g−1), and a KOHactivated carbon ACK (SBET = 2180 m2 g−1). The electrolytes were: aqueous 1 M Li 2 SO 4 (≥ 99% puriss p.a., Sigma-Aldrich) and 0.5 M KI (99.5%, Avantor); non-aqueous 1 M tetraethylammonium tetrafluoroborate (TEABF4, ≥ 99%, SigmaAldrich) in acetonitrile (AN, 99.8%, Sigma-Aldrich) and 1 M NaClO4

22

A. Laheäär et al. / Electrochemistry Communications 60 (2015) 21–25

Table 1 Comparison of SC characteristics derived from different calculation methods. An average electrode capacitance value is expressed, if not indicated otherwise. Electrolyte

Electrode material

Umax/ V

Cel,GD/ F g−1

Cel,int/D/ F g−1

Qel,D/ mAh g−1

Es,GD/ Wh kg−1

Es,int/D/ Wh kg−1

ηt/ %

ηE/ %

EMImTFSI

YP-80F

BMImBF4

YP-80F

1 M Li2SO4

ACK

0.5 M KI

ACK

1 M TEABF4 in AN 1 M NaClO4 in EC:DMC

ACK C(Mo2C)

3.2 (Initial) 3.2 (100 h)a 3.2 (Initial) 3.2 (400 h)a 1.2 1.6 1.2 1.6 2.5 3.0 3.0 (5000)b

108 126 107 103 44 (cell)c 176 54 (cell)c 215 83 (cell)c 179 (−)d 126 (cell)c 273 (−)d 146 116 99

113 112 109 95 43 (cell)c 172 47 (cell)c 187 82 (cell)c 175 (−)d 93 (cell)c 194 (−)d 142 110 94

– – – – 14.5 24 27.5 56 – – –

38 45 38 36.5 8.8 19 16.5 44.5 31.5 27 23

40 40 39 33.5 8.6 16.5 16.5 33.0 31 26 22

97 81 100 100 97 96 97 91 98 100 99

89 59 90 82 90 71 86 52 93 96 95

a b c d

Time of floating test at 3.2 V. Number of galvanostatic cycles between 1.5 V and 3.0 V. SC system specific capacitance (per total mass of two electrodes). (−) electrode specific capacitance.

(98%, Sigma-Aldrich) in a 1:1 volume mixture of ethylene carbonate (EC, 99%, Sigma-Aldrich) and dimethyl carbonate (DMC, 99%, Sigma-Aldrich); and ionic liquids (ILs) 1-ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)imide (EMImTFSI, 99.5%, Solvionic) and 1-butyl-3-methylimidazolium tetrafluoroborate (BMImBF4, 99%, Iolitec). The electrochemical measurements were performed with a VMP3 multichannel potentiostat/galvanostat (Bio-Logic) in two-electrode PTFE Swagelok-type test cells. In order to monitor separately the potential of individual electrodes, the cells with KI and Li2SO4 electrolytes were equipped with Ag/Ag+ and saturated calomel reference electrode (RE), respectively.

In case of non-linear GC/GD characteristics, the discharge energy should be calculated by integrating, i.e. finding the area under the GD curve (or charge energy Eint/C from GC):

3. Results and discussion

Subsequently, an average discharge capacitance Cint/D can be extracted from Eint/D, which was referred to as “claimed capacitance” in Ref. [4]. Obviously, when C is a function of U, such average Cint/D value does not reflect the behavior at various levels of capacitor charging/discharging, as does CGD for EDLC linear trend. Nevertheless, such value is much more realistic than the one obtained by entering the extreme U values in Eq. (1), as done in many publications. Therefore, opposite to common practice, we suggest in a first step to determine the appropriate discharge energy directly from the GD curve, and then to extract the system capacitance by using the energy-capacitance relation in Eq. (3):

3.1. Calculation methods The capacitance (F) of an ideal EDLC is calculated from the slope of the galvanostatic charge/discharge (GC/GD) curves:

C GC=GD ¼

IΔt ; ΔU

ð1Þ

where I is the applied current (A), Δt is the discharge or charge duration (s) and ΔU is the change in cell potential (V) corrected from the ohmic drop. Thus, the specific discharge capacitance (F g− 1) of a single electrode is:

C el;GD ¼

2C GD ; 0:5mel

ð2Þ

where mel (g) is the total mass of electrodes. The SC discharge energy EGD (W s) and specific energy Es,GD (Wh kg−1) can then be calculated:

EGD ¼

C GD U 2max EGD ; Es;GD ¼ : 2 mel 3:6

ð3Þ

It is crucial to keep in mind that Eqs. (1) and (3) should only be applied for ideally linear GC/GD curves. For SCs with storage based on faradaic processes (redox electrolytes, hydrogen electrosorption, metal oxides, conducting polymers, etc. [3]), as well as for EDLCs with any degradation process, non-linearity appears in the GC/GD curves, and different calculations must be adapted to obtain adequate information.

t ðU Zmin Þ

Eint=D ¼ I

U ðt Þdt;

ð4Þ

t ðU max Þ

and the corresponding specific energy: Es;int=D ¼

C int=D ¼

Eint=D : mel 3:6

2E int=D U 2max

ð5Þ

:

ð6Þ

An average electrode specific capacitance (Cel,int/D) can be estimated from Cint/D by applying Eq. (2), but only for SCs with electrodes displaying similar capacitive behavior. Since the integration method in Eq. (4) involves the extreme potential limits, the results obtained are somewhat influenced by the chosen potential range. Therefore, cyclic voltammograms must be recorded in parallel with GC/GD curves in order to evaluate the stability potential limits, as well as to differentiate between battery-like and pseudocapacitive faradaic systems (well explained by examples in Ref. [3]). A system (or electrode) exhibiting battery-like GC/GD curves should be described by specific discharge capacity QD (or charge capacity QC) in mAh g−1: QD ¼

It D mel  3:6

 QC ¼

 It C ; mel  3:6

where tD and tC are the discharge and charge durations (s).

ð7Þ

A. Laheäär et al. / Electrochemistry Communications 60 (2015) 21–25

The efficiency of SCs is very often presented as coulombic efficiency (ηt), determined by the ratio of QD to QC, which according to (7) is: ηt ¼

tD : tC

ð8Þ

However, for non-ideal GC/GD, ηt and energy efficiency (ηE) can differ significantly. Therefore, efficiency should be evaluated from: ηE ¼

Eint=D : Eint=C

ð9Þ

Further in this manuscript, considering faradaic systems or aging experiments on EDLCs, we demonstrate how it is important to evaluate efficiency by different methods in order to assess the reversibility of SCs. 3.2. Capacitance and efficiency change during EDLC aging in ILs Imidazolium-IL based capacitors were investigated by floating, i.e. accelerated aging by holding the cell for prolonged periods at Umax, with periodic evaluation of cell's state-of-health by few galvanostatic cycles [6]. It is generally accepted that a SC is out of service when capacitance is decreased by 20% or resistance increased by 100% [1,7]. The failure by resistance was established for EMImTFSI based EDLCs after 100 h floating at Umax = 3.2 V. Surprisingly, contrary to the common trend of capacitance decrease during aging, an increase in Cel,GD values was evaluated from Eq. (2), thus also an increase in Es,GD values (Table 1). However, the GC/GD curves slightly deviate from linearity after few hours of floating, being more noticeable for the charging process (Fig. 1). Therefore, we decided to evaluate the characteristics of this system by integration (Eqs. (4) to (6)). As shown in Table 1, and as commonly expected, the discharge capacitance calculated by such way does not increase. However, the Es,int/C values increase continuously during floating, leading to very poor ηE of around 60% (colored areas in Fig. 1a). Such behavior is related to faradaic decomposition reactions as Umax is approached. By contrast, the EDLCs based on the more stable BMImBF4 IL, display close to linear GC/GD curves throughout 400 h floating at 3.2 V (Fig. 1b), leading to smaller differences between Cel,GD and Cel,int/D values and higher final ηE of 82% (Table 1). It is worth to notice that, before floating, ηE was ~ 90% for both IL systems, differently from the ηt of 97% for EMImTFSI and even 100% for BMImBF4.

23

hybrid SCs with redox active electrolytes. In the first case, neutral aqueous electrolytes (example in Fig. 2a and b with Li2SO4) allow the voltage of AC/AC capacitors to be enhanced up to 1.5–1.8 V, due to a high overpotential for di-hydrogen evolution resulting from pH increase in the pores of the negative electrode [8–10]. In the second case, advantage is taken of a very small potential change of one electrode during charge/discharge, owing to the extremely high capacity related with fast reversible redox reactions (example in Fig. 2c and d with the positive electrode employing iodide reactions in 0.5 M KI), so that the cell voltage profile is practically determined by the EDL type electrode [11–15]. Since C+ N N C− in the latter case, and considering the relation 1/Ccell = 1/C+ + 1/C−, the cell capacitance (and energy) in iodide solution should be roughly two times higher than in an EDLC with C+ = C−. The Es,int/D values at the same Umax in Table 1 prove such concept, where Cel,int/D in 1 M Li2SO4 is practically equal to Cel,int/D of the negative (EDL type) electrode in the 0.5 M KI based SC, but the cell specific capacitance is two times higher in the latter system. Notwithstanding, capacity values were preferred, instead of presenting a meaningless value of, for example, 3500 F g−1 for the iodide redox electrode. They indicate that Qel,D for the 0.5 M KI based SC is ~2 times higher than that for the non-redox system, being in accordance with the estimated energy densities (Table 1). Owing to the linearity of the GD curves for both aqueous systems at Umax = 1.2 V (Fig. 2), Cel,GD ≈ Cel,int/D and Es,GD ≈ Es,int/D (Table 1). However, the ηE values (86–90%) are once more much lower than ηt (97%). At Umax = 1.6 V, faradaic hydrogen electrosorption (or even H2 evolution in the KI system) takes place in addition to EDL charging, and it induces non-linearity of the GC/GD curves (Fig. 2). The corresponding data in Table 1 indicate that, depending on the calculation method, very different characteristics are obtained, and considering such system identical to an EDLC results in dramatic performance overestimation. The energy efficiency is remarkably more sensitive to the maximum applied potential than the coulombic efficiency. According to Ref. [16], the low ηE ≈ 50% at 1.6 V for the KI system could be due to a blockage of the hydrogen adsorption sites by the iodide reaction products (iodine and iodates), leading to irreversible di-hydrogen evolution. Misleading coulombic efficiency values up to 100%, without evaluation of energy efficiency, have also been reported for other faradaic systems (conducting polymer, functionalized carbon with redox electrolyte, etc.) [17–20].

3.4. Is the efficiency of EDLCs close to 100%? 3.3. Performance overestimation of capacitors based on redox process For non-linear GC/GD behavior, 2-electrode cells equipped with RE should be applied to evaluate the real capacitance/capacity and efficiency of each electrode from its GC/GD plot. Redox/faradaic processes at the AC/aqueous electrolyte interface are involved either at the negative electrode when hydrogen is chemisorbed in the carbon porosity or in

Efficiency exceeding 95% is generally claimed for EDLCs based on porous carbon and 1 M TEABF4 in AN (or PC). Table 1 demonstrates that ηt at Umax = 2.5 V is indeed 98%, however ηE is somewhat less—93%. The same conclusion was derived for EDLCs operating at U = 3.0 V using 1 M NaClO4 in EC:DMC—ηt ≈ 100% and ηE ≈ 95% (GC/GD curves published in [21]). Thus, even for EDLCs with nearly ideal GC/GD profiles,

Fig. 1. GC/GD characteristics (at 0.1 A g−1 per system) for freshly prepared (initial) SCs with YP-80F electrodes and after floating at 3.2 V for (a) 100 h in EMImTFSI and (b) 400 h in BMImBF4. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

24

A. Laheäär et al. / Electrochemistry Communications 60 (2015) 21–25

Fig. 2. GC/GD characteristics (at 0.1 A g−1 per system) of cell (with RE) as well as (+) and (−) ACK electrodes in (a,b) 1 M Li2SO4 and (c,d) 0.5 M KI at Umax = 1.2 V (a,c) and Umax = 1.6 V (b,d).

there is a small discrepancy in the estimated performance when applying different calculation methods. Such reduced energy efficiency is at least partly due to heat dissipation, as demonstrated by calorimetric measurements during SC galvanostatic cycling [22,23]. Hence, in the typical EDL capacitors (organic medium, IL and Li2SO4 below electrolyte decomposition voltage), the main reason for efficiency loss would be this dissipated heat. In the cases involving faradaic charge storage, the low energy efficiency (as low as 50%) is essentially related to the irreversibility of faradaic processes. Also decomposition products, which are formed in the electrodes, block partly the porosity and change the EDL characteristics.

4. Conclusions Assuming that all types of supercapacitors behave similarly to electrical double-layer capacitors, whatever the electrode/electrolyte couple or cycling conditions, may result in erroneous interpretation of data. The fare exploitation of non-ideal capacitance characteristics (non-linear potential-time dependence at constant current) requires mathematical processing of data based on energy calculated from integration of galvanostatic curves. By this way, it could be demonstrated that presenting only EDLC coulombic efficiencies close to 100%, whilst not evaluating energy efficiency, results in overestimated description. Similarly, for systems involving faradaic charge storage, the efficiency is inferior to 80% or even down to 50%, instead of misleading coulombic efficiency of 90–95%. Therefore, the recommendation of this paper is that, for all future reports on SCs (including EDLCs), energy efficiency should be reported in parallel with coulombic efficiency for more realistic characterization. Moreover, specific capacitance values (F g−1) should not be presented for faradaic electrodes having battery-like behavior, as these processes do not have “capacitance”. The comparison of such electrodes/devices should be preferably done on the basis of device energy density or capacity.

Conflict of interest The authors declare that there are no conflict of interest.

Acknowledgments The Foundation for Polish Science is acknowledged for funding the ECOLCAP Project realized within the WELCOME/2010-4/1 Program, co-financed from European Union Regional Development Fund. The authors thank Kuraray for kindly providing the YP-80F carbon.

References [1] F. Béguin, E. Frackowiak, Supercapacitors: Materials, Systems, and Applications. Wiley-VCH, Weinheim, 2013. [2] S. Roldan, D. Barreda, M. Granda, R. Menendez, R. Santamaria, C. Blanco, An approach to classification and capacitance expressions in electrochemical capacitors technology, Phys. Chem. Chem. Phys. 17 (2015) 1084–1092. [3] T. Brousse, D. Belanger, J.W. Long, To be or not to be pseudocapacitive? J. Electrochem. Soc. 162 (2015) A5185–A5189. [4] B. Akinwolemiwa, C. Peng, G.Z. Chen, Redox electrolytes in supercapacitors, J. Electrochem. Soc. 162 (2015) A5054–A5059. [5] A. Jänes, T. Thomberg, H. Kurig, E. Lust, Nanoscale fine-tuning of porosity of carbidederived carbon prepared from molybdenum carbide, Carbon 47 (2009) 23–29. [6] D. Weingarth, A. Foelske-Schmitz, R. Kötz, Cycle versus voltage hold—which is the better stability test for electrochemical double layer capacitors? J. Power Sources 225 (2013) 84–88. [7] Application note—125 V and 390 V modules—©2007 Maxwell Technologies Inc, www.maxwell.com. [8] Q. Gao, L. Demarconnay, E. Raymundo-Pinero, F. Béguin, Exploring the large voltage range of carbon/carbon supercapacitors in aqueous lithium sulfate electrolyte, Energy Environ. Sci. 5 (2012) 9611–9617. [9] P. Ratajczak, K. Jurewicz, P. Skowron, Q. Abbas, F. Béguin, Effect of accelerated ageing on the performance of high voltage carbon/carbon electrochemical capacitors in salt aqueous electrolyte, Electrochim. Acta 130 (2014) 344–350. [10] P. Ratajczak, K. Jurewicz, F. Béguin, Factors contributing to ageing of high voltage carbon/carbon supercapacitors in salt aqueous electrolyte, J. Appl. Electrochem. 44 (2014) 475–480. [11] G. Lota, E. Frackowiak, Striking capacitance of carbon/iodide interface, Electrochem. Commun. 11 (2009) 87–90.

A. Laheäär et al. / Electrochemistry Communications 60 (2015) 21–25 [12] S. Roldan, M. Granda, R. Menendez, R. Santamarıa, C. Blanco, Mechanisms of energy storage in carbon-based supercapacitors modified with a quinoid redox-active electrolyte, J. Phys. Chem. C 115 (2011) 17606–17611. [13] V. Khomenko, E. Raymundo-Piñero, F. Béguin, High-energy density graphite/AC capacitor in organic electrolyte, J. Power Sources 177 (2008) 643–651. [14] K. Naoi, P. Simon, New materials and new configurations for advanced electrochemical capacitors, Interface 17 (2008) 34–37. [15] T. Aida, K. Yamada, M. Morita, An advanced hybrid electrochemical capacitor that uses a wide potential range at the positive electrode, Electrochem. Solid-State Lett. 9 (2006) A534–A536. [16] K. Fic, M. Meller, E. Frackowiak, Interfacial redox phenomena for enhanced aqueous supercapacitors, J. Electrochem. Soc. 162 (2015) A5140–A5147. [17] S. Bose, T. Kuila, A.K. Mishra, R. Rajasekar, N.H. Kim, J.H. Lee, Carbon-based nanostructured materials and their composites as supercapacitor electrodes, J. Mater. Chem. 22 (2012) 767–784. [18] S.A. Hashmi, A. Kumar, S.K. Tripathi, Investigations on electrochemical supercapacitors using polypyrrole redox electrodes and PMMA based gel electrolytes, Eur. Polym. J. 41 (2005) 1373–1379.

25

[19] M. Mastragostino, C. Arbizzani, R. Paraventi, F. Soavi, A. Zanelli, Polymer based supercapacitors: selection of material and cell design, in: G.A. Nazri, M. Thackeray, T. Ohzuku (Eds.), Intercalation Compounds for Battery Materials, The Electrochemical Society Proceedings Series, Pennington, New Jersey 2000, pp. 416–421. [20] L.-Q. Mai, A. Minhas-Khan, X. Tian, K.M. Hercule, Y.-L. Zhao, X. Lin, X. Xu, Synergistic interaction between redox-active electrolyte and binder-free functionalized carbon for ultrahigh supercapacitor performance, Nat. Commun. 4 (2013) 2923–2930. [21] A. Laheäär, A. Jänes, E. Lust, NaClO4 and NaPF6 as potential non-aqueous electrolyte salts for electrical double layer capacitor application, Electrochim. Acta 82 (2012) 309–313. [22] C. Pascot, Y. Dandeville, Y. Scudeller, P. Guillemet, T. Brousse, Calorimetric measurement of the heat generated by a Double-Layer Capacitor cell under cycling, Thermochim. Acta 510 (2010) 53–60. [23] Y. Dandeville, P. Guillemet, Y. Scudeller, O. Crosnier, L. Athouel, T. Brousse, Measuring time-dependent heat profiles of aqueous electrochemical capacitors under cycling, Thermochim. Acta 526 (2011) 1–8.