study on specific effects of high frequency ripple...

Microelectronics Reliability xxx (2015) xxx–xxx

MR-11575; No of Pages 5

Contents lists available at ScienceDirect

Microelectronics Reliability

j ourna l homepage: www.e lsev ie r .com/ locate /mr

Study on specific effects of high frequency ripple currents and temperature onsupercapacitors ageing

R. German a,⁎, A. Sari a, P. Venet a, O. Briat b, J.-M. Vinassa b

a Université de Lyon, Université Lyon 1, Laboratoire Ampère, UMR CNRS 5005, Villeurbanne F-69622, Franceb Université de Bordeaux, IMS, UMR CNRS 5218, Talence F-33405, France

⁎ Corresponding author.E-mail address: [email protected] (R. Ge

http://dx.doi.org/10.1016/j.microrel.2015.06.0260026-2714/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article as: German R, et al, StuMicroelectronics Reliability (2015), http://dx

a b s t r a c t
a r t i c l e i n f o
Article history:Received 25 May 2015Received in revised form 19 June 2015Accepted 20 June 2015Available online xxxx

Keywords:Power devicesEnergy storage systemsSupercapacitorsReliabilityAgeing

Although supercapacitors can store less energy than batteries, they are particularly appealing for hybrid vehicleapplications (for example, for braking energy recovery, or stop and start systems supply…). As power networksin hybrid vehicle experience high frequency ripple currents due to power electronics components such as staticconverters, experiments are done to study the impact of high frequency ripple currents on supercapacitor ageing.The results suggest that high frequency ripple currents do not create new ageingmechanisms. A complementaryageing test is also carried out to study the effect of temperature on supercapacitor ageing at very low voltage. Itclearly shows that a significant part of increase in supercapacitor resistance is only linked to temperature, insteadof the combined effect of temperature and voltage as assumed by literature for supercapacitor ageing. Whereastests demonstrate that capacitance loss through time is linked (for themost part) to the combined effect of voltageand temperature. Hence, this paper shows that resistance increase is only partially linked to capacitance loss.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Supercapacitors (SCs) [1], also called Electrochemical Double LayerCapacitors [2] or Ultracapacitors [3,4] are energy storage systems(ESSs) capable of delivering higher power than batteries over a widerrange of temperature [5], and with a longer lifetime, than batteries [6].These characteristics make SCs desirable for high cyclability/highpower applications in railway [7] and electrical traction systems,such as personal hybrid vehicles (HVs) [8], or as a supplement forbatteries [9].

Power electronics in HVs power network (such as static converters)generate high frequency ripple currents. To understand their effect onSCs ageing, tests are performed for representative frequencies existingin different applications. Selected for testing are commercial 3000-Farad SCs from 3 different manufacturers built with the most commontechnology (carbon/carbon electrodes and organic acetonitrile/TEABF4electrolyte).

For most of the studied frequencies, (except for 1 kHz test which hasbeen perturbed by voltage regulation problems), HF ripple current doesnot influence SC ageing.

To demonstrate that 1 kHz should not be an exception to theobservation above, a second test is performed, but at reduced voltage.In this test, HF 1 kHz ripple currents have no effect on SC ageing. This

rman).

dy on specific effects of high.doi.org/10.1016/j.microrel.2

test also allows us to draw new conclusions. SC resistance increase ap-pears to be only linked to temperature, which is different from capaci-tance loss mechanisms that need both voltage and temperature.

2. SC operational mode and ageing

SCs use electrostatic effect between ions and electric charges(double layer effect) to store energy (go to Fig. 7). The electrodes areporous to maximize the storage surface (thus the capacitance), and tominimize the equivalent series resistance (ESR). As the SCs use ionsand porous electrodes for energy storage, they are only able to storeenergy at low frequency (LF) as shown in Fig. 1.

SC capacitance (C100mHz) is calculated in Eq. (1) for a 100 mHzfrequency and represents the energy stored in LF (at 100 mHz thecapacitance is near the maximum as shown in Fig. 1). The 100 mHzESR (R100mHz), representing the cumulative effect of connections,conduction and porosity on current circulation, is calculated in Eq. (2).

C100mHz ¼ −12 � π � f � Im ZSCð Þ

��f¼100mHz

ð1Þ

R100mHz ¼ Re ZSC 2 � π � fð Þð Þj f¼100mHz ð2Þ

Where ZSC is the SC impedance and f the frequency of the electricalsignal. Ageing phenomena for SCs lead to a decrease in capacitanceand an increase in ESR. They are the result of reactions between parasiticchemical groups at the electrode surface [10] and electrolyte [11].

frequency ripple currents and temperature on supercapacitors ageing,015.06.026

http://dx.doi.org/10.1016/j.microrel.2015.06.026

mailto:[email protected]


http://www.sciencedirect.com/science/journal/

www.elsevier.com/locate/mr


Fig. 1. Evolution of SC capacitance with frequency.

Fig. 3. Thermocouple location on the SC package.

2 R. German et al. / Microelectronics Reliability xxx (2015) xxx–xxx

Literature considers that those reactions are accelerated by tempera-ture, voltage and current rate (in the case of cycle ageing) [12].

Table 1Electric characteristics of different manufacturer acetonitrile/active carbon commercial3000 F SC.

Electrode Activated carbon

Electrolyte Acetonitrile

Capacitance 3000 F

Manufacturer A, B, CEquivalent series resistance ESR (mΩ) 0.20 b ESR b 0.29Nominal voltage UN (V) 2.7 ≤ UN ≤ 2.8Limit impulse current b1 s (A) ≈2000 A

3. Experiment setup

Fig. 2 presents the test bench developed for SC ageing tests. Thefloating (i.e. constant) constraints (USC, T) are applied by independentDC power supplies and a temperature chamber. The ripple current(IAC) is applied to the stacked SCs by a programmable AC current supplyand the DC supplies are protected from ripple current by coils (L1…Ln).The temperature of SCs during ageing is monitored by a thermocoupleplaced on the middle of the component packaging (see Fig. 3).

We tested 33 SCs from three manufacturers (A, B, C) to get a repre-sentative sample of commercial high capacitance ACN/TEABF4 SCs [13].The tested SCs are equivalent in terms of technology and performancesas shown in Table 1.

The tested SCs are high capacitance (3000 F), high power(ESR b0.29 mΩ), and low voltage elements (b2.8 V to minimize thesolvent decomposition rate), which are typically designed for hybridvehicle use. As presented in Table 2, three components of each man-ufacturer are aged at 2.8 V, 60 °C floating constraints (i.e. constantvoltage and temperature).

For a second batch of SCs (frommanufacturers A, B and C), sinewaveripple currents are added to the floating constraints on two SCs bymanufacturer and by investigated ripple current (see Table 2). Threeparticular frequencies are investigated for 2.8 V, 60 °C floatingconstraints with a 12 ARMS value for ripple current. The RMS value hasbeen estimated by the industrial and academic partners of Supercal pro-ject (see Acknowledgements) based on the usual ripple rates of DC–DCconverters used in hybrid vehicles. The 10 kHz frequency is in the sameorder of magnitude of DC–DC converters recorded in HVs powernetwork. The 1 kHz frequency approximately corresponds to the mini-mum value of ZSC for the tested SCs. The 100 Hz frequency correspondsto the limit of capacitive zone of tested SCs [14] (see Fig. 1) or a full waverectified 50 Hz signal (for example when SCs are charging).

After observing the result from the above test, a second ageing test isperformed for 1 kHz ripple current at 0.1 V.

Fig. 2. Scheme of the experimental test bench.

Please cite this article as: German R, et al, Study on specific effects of highMicroelectronics Reliability (2015), http://dx.doi.org/10.1016/j.microrel.2

At regular intervals, SCs are disconnected from the DC and AC sup-plies and an impedance spectroscopy is carried out on each componentfrom 10 kHz to 10 mHz.

4. Results

4.1. Effect of ripple current frequency [14,15]

Fig. 4 presents the evolution of C100mHz and R100mHz as a function ofageing time. The ageing tests take up to 6600 h (9months) and lead to aminimum capacitance decrease of 30%. As SCs are considered not usableafter 20% for capacitance drop [13], the results are very representative ofSCs lifetime.

For 100 Hz and 10 kHz, the impact of ripple current on SCs is notevidently different from the (2.8 V, 60 °C) floating ageing results. Thefrequency also seems not to be a significant ageing acceleration factor,as the results are the same for 100 Hz and 10 kHz ripple current. Theripple current does not contribute to an increase of SC temperature asshown in Fig. 5. This is because the SCs are high current components(see Table 1) designed for a 15 °C joule effect self-heating when a130 ARMS current is applied [16]. As the self-heating is proportional tothe square of the current, the self-heating from 12 ARMS current shouldbe negligible.

Fig. 5 presents the evolution of SCs temperature during ageing tests.For 100Hz and10 kHz ripple current, the temperature does not increaseat all before 5000 h. The slight increase of temperature (b1.5 °C)observed for manufacturers A and B between 5000 h and 7000 h (for10 kHz ageing test) can be related to the change in slope of the SCR100mHz parameter after 5000 h for 10 kHz ageing test (see Fig. 4)possibly indicating an effect of overageing on connections (at thisstage the SCs are in extreme capacitance loss state (N−20%)). This neg-ligible rise of temperature is a confirmation of the results presented in

Max operating temperature Tmax (°C) 60 b Tmax b 65Cell weight M (g) 510 b M b 650Mass energy WM (kJ/kg) 17.8 b WM b 21.5Peak mass power PM b 1 s (kW/kg) 8.60 b PM b 10.6

Table 2Experiments setup.

Floating constraints USC = 2.8 VT = 60 °C

USC = 0.1 VT = 60 °C

RMS currentRipple (IAC)

0 A 12 A100 Hz

12 A1 kHz

12 A10 kHz

12 A1 kHz

Number of tested elementsA, B, C

3, 3, 3 2, 2, 2 2, 2, 2 2, 2, 2 2, 2, 2



Fig. 4. Evolution of C100mHz and R100mHz with ageing time.

3R. German et al. / Microelectronics Reliability xxx (2015) xxx–xxx

Fig. 4 (100 Hz and 10 kHz). It tends to show that the HF current ripplesare not a factor causing ageing acceleration for SCs.

In comparison, the results obtained for 1 kHz current ripple areslightly different from the floating ageing results (for capacitanceC100mHz) and even more different for R100mHz [15]. Those results couldbe due to a particular effect of 1 kHz frequency on gas production duringageing such as suggested in [17] or to the ageing test disturbancecaused by two voltage breakdowns (during the 1 kHz test the voltageregulation system broke down twice as shown in Fig. 4).

During the breakdowns voltage was uncontrolled and unrecordedleading to a succession of overvoltage and undervoltage for severalhours. Before the first breakdown the SCs aged at 1 kHz are strictlyageing in the sameway as floating ageing test. Manufacturer C elementswere stopped because they started to leak. The second breakdown isaccompanied by a sudden increase in R100mHz. The temperature alsoincreases significantly by 3 °C.

The increase of R100mHz cannot explain the temperature increaseby joule effect (see manufacturer data on selfheating [16]). That seemsto indicate that the voltage breakdowns lead to a modification ofelectrolyte and creates a few exoenergetic reactions in the electrolyte.

To be sure that the differences observed between the 1 kHz and thefloating ageing tests are caused by the two breakdowns but not theparticular effect of 1 kHz current ripple, we decide to observe the effect,of the 1 kHz current ripple again, but this time at very low voltage tominimize ageing related to floating constraints.

4.2. Effect of temperature on SC ageing (at low voltage)

To test the effect of ripple current at 1 kHz, a second ageing test ofone year (8700 h) is carried out. The SCs of the three manufacturers


are placed at 60 °C in a thermal oven andmaintained at very low voltage(+0.1 V to avoid negative voltage). As it is commonly noted bymanufacturer, and verified in literature [8], an increase of theSC voltage by 200 mV results in an increase of ageing kinetic by afactor 2.

Thus, at 100 mV, the ageing kinetic due to voltage is negligible(104 factors smaller compared to a 2.8 V ageing at the same temper-ature condition). Thus, the capacitance loss and the ESR gain must bevery small.

The test consists of two phases as shown in Fig. 6. For phase 1,1 kHz/12 ARMS ripple current is applied to the SCs. Then, for phase 2(just before 6000 h), the ripple current is stopped. Then, during phase2, SCs are only subjected to 0.1 V, 60 °C floating conditions. Fig. 6presents the evolution of C100mHz and R100mHz for this test. We cannotice that the capacitance and the resistance evolution are notaffected when the ripple current stops for phase 2. This clearlyshows, that ripple currents have no particular effect, on the testedSCs ageing (as long as their RMS value small enough to not createSC major self-heating).

C100mHz is hardly (b2%) affected by the 1 year (8700 h) ageing test at0.1 V. Thus, storing short-circuited SCs, even at high temperature,will have a very small impact on their capacitance.

The evolution of R100mHz is quite different. At 6000 h, by comparingwith the R100mHz increase with 2.8 V, 60 °C floating ageing results,we notice that the components of the manufacturer A and B gainrespectively +40% and +10% for R100mHz (@ 6000 h for 0.1 V, 60 °C),which represents 50% and 25% of the increase of R100mHz (@6000 h for2.8 V, 60 °C). From that, we can conclude that temperature plays animportant part in causing resistance increase (even when voltageis negligible).



Fig. 5. Evolution of temperature for 100 Hz, 1 kHz and 10 kHz current ripple.

Fig. 7.Movement of charges for HF AC current.

4 R. German et al. / Microelectronics Reliability xxx (2015) xxx–xxx

This phenomenon is independent of the 1 kHz ripple current, as animportant increase of R100mHz happenswith, andwithout, ripple current(see Fig. 6).

This significant increase cannot be explained by the accelerationfactor observed on the capacitance loss.

Consequently, we can deduce that the increase of R100mHz, iscaused by several mechanisms: A part of the increase of R100mHz isinfluenced by the chemical reactions linked with the capacitancedecrease (i.e. gas creation caused by reaction between parasitic surfacegroup and surface group and electrolyte). Another part of the R100mHz

increase is linked with totally different phenomenon linked only totemperature. As this part of resistance increase is not linked to voltage,

Fig. 6. Evolution of C100 mHz and R100 mHz.


which is required for capacitance ageing, we can conclude that thesupplementary effects causing R100mHz increase are not linked withcapacitance ageing (thus they probably take place outside theelectrodes porosity).

We can also notice that the supplementary mechanisms for R100mHz

ageing are not necessarily present for all SCs manufacturers. Forexample, for manufacturer C, R100mHz stays constant for the low voltageageing test. It indicates that the increase of resistance ismainly linked tothe capacitance loss for manufacturer C.

Thus, the supplementary causes of R100mHz increase, are linked withporosity engineering factors, which tend to reduce the charge flow,without any relation to electrode storage surface.

The first possible cause can be the degradation of the binding agentthat sticks active carbon electrode and collector together. This makesthe contact area smaller. The second possible cause could be thatsome parasitic chemicals inside the electrolyte dislocate without otherreactant (zero order reaction), or react togetherwith temperature effectleading to a degradation of electrolyte conductivity. Finally, the separa-tor itself can also become less conductive for ions due to ageing, becauseof fouling and/or slow decompositionmechanisms (the SCs separator isjust simply built of paper).

4.3. Interpretation of HF ripple current transparency

A physical interpretation is given in Fig. 7. As described in part 2, SCsare relatively low frequency ESSs. At the ageing test frequencies (be-tween 100 Hz and 10 kHz), we can deduce that ions are not fast enoughto compensate the supplementary electric charges from the AC current(the capacitance is close to 0 F for this frequency range as shown inFig. 1). Supplementary electric charges from one electrode negatewith the supplementary charges from the other electrode. Thus, thereis no supplementary formation of double layer and thus no supplemen-tary ageing reactions between electrolyte and electrodes.

This explains why SCs are not affected by high frequency currentripples.

5. Conclusion

This article presents the possible effect of HF current ripple and theparticular effect of temperature on SC ageing kinetic.

33 SCs from 3 different manufacturers are aged up to 12 monthswith or without ripple current. Three frequencies (100 Hz, 1 kHz and10 kHz) are investigated.

As these SCs are designed for around 120 ARMS current use and as themaximum ripple induced byDC–DC converters is estimated to be 10% ofthe mean current, the current ripple for one component is fixed to12 ARMS.



5R. German et al. / Microelectronics Reliability xxx (2015) xxx–xxx

The evolution of the electric parameters of SCs (C100mHz andR100mHz) is monitored with regular impedance spectroscopies. Thetemperature of SCs is also monitored during ageing.

The tests reveal that HF current ripples do not have any significanteffect on SC ageing. Consequently, we can conclude that the supplemen-tary electric charges from AC current do not participate in double layerformation, because the frequency is too high for ions, to compensatethe supplementary charges brought by ripple current. Thus, HF ripplecurrents do not impact SCs, because no new zone of double layer ageingis created.

These conclusions have a great appeal to SC integrators, as theywon't have to use costly coils to protect SCs from HF ripple currents(from domestic appliance (~100Hz) to high frequency static converters(~10 kHz)).

The complementary test shows that SC capacitance is hardly affectedunder low voltage/high temperature ageing test whereas theirresistance can increase a lot. That shows that the ageing mechanismsaffecting SC capacitance and SC ESR are not identical.

Acknowledgement

French national research agency (ANR) is supporting these researchworks which are part of the Supercal project. Supercal is a projectheld by IMS Bordeaux which combines three academic laboratories(Ampère (Lyon), IMS (Bordeaux), French institute of sciences andtechnology for transport, development and networks (IFSTTAR)) andmanufacturers from transportation and energy storage devices sectors(PSA Peugeot Citroën, Valeo, Batscap).

References

[1] Ayadi M, Briat O, Lallemand R, Eddahech A, German R, Coquery G, Vinassa JM.Description of supercapacitor performance degradation rate during thermal cyclingunder constant voltage ageing test. Microelectron Reliab Sept.–Oct. 2014;54(9–10):1944–8.


[2] Kang J, Wen J, Jayaram SH, Yu A, Wang X. Development of an equivalent circuitmodel for electrochemical double layer capacitors (EDLCs) with distinct electrolytes.Electrochim Acta Jan. 2014;115:587–98.

[3] Brouji HE, Briat O, Vinassa J-M, Bertrand N, Woirgard E. Comparison betweenchanges of ultracapacitors model parameters during calendar life and power cyclingageing tests. Microelectron Reliab 2008;48(8):1473–8.

[4] Zhang L, Wang Z, Hu X, Sun F, Dorrell DG. A comparative study of equivalentcircuit models of ultracapacitors for electric vehicles. J Power Sources Jan.2015;274:899–906.

[5] Kang J, Jayaram SH, Rawlins J, Wen J. Characterization of thermal behaviors ofelectrochemical double layer capacitors (EDLCs) with aqueous and organicelectrolytes. Electrochim Acta Oct. 2014;144:200–10.

[6] Ongaro F, Saggini S, Mattavelli P. Li-ion battery-supercapacitor hybrid storagesystem for a long lifetime, photovoltaic-based wireless sensor network. IEEE TransPower Electron Sep. 2012;27(9):3944–52.

[7] Wang W, Cheng M, Wang Y, Zhang B, Zhu Y, Ding S, Chen W. A novel energymanagement strategy of onboard supercapacitor for subway applications withpermanent-magnet traction system. IEEE Trans Veh Technol Jul. 2014;63(6):2578–88.

[8] German R, Sari A, Venet P, Ayadi M, Briat O, Vinassa J-M. Prediction ofsupercapacitors floating ageing with surface electrode interface based ageing law.Microelectron Reliab Sept.–Oct. 2014;54(9–10):1813–7.

[9] Lahyani A, Venet P, Guermazi A, Troudi A. Battery/supercapacitors combination inuninterruptible power supply (UPS). IEEE Trans Power Electron 2013;28(4):1509–22.

[10] Toles CA, Marshall WE, Johns MM. Surface functional groups on acid-activatednutshell carbons. Carbon 1999;37(8):1207–14.

[11] Azais P, Duclaux L, Florian P, Massiot D. Causes of supercapacitors ageing in organicelectrolyte. J Power Sources 2007;171:1046–53.

[12] Chaari R, Briat O, Deletage JY, Woirgard E, Vinassa J-M. How supercapacitors reachend of life criteria during calendar life and power cycling tests. Microelectron Reliab2011;51(911):1976–9.

[13] M. technologies. ProductGuide,Maxwell Technologies BOOSTCAPUltracapacitors; 2009.[14] German R, Sari A, Venet P, Briat O, Vinassa J-M. Study of Static Converters related

Ripple Currents Effects on Supercapacitors Ageing within DC Networks. Presentedat the Industrial Electronics (ISIE), 2014 IEEE 24rd International Symposium on,Rio de Janeiro; 2015.

[15] German R, Briat O, Sari A, Venet P, Ayadi M, Zitouni Y, Vinassa J-M. Impact of highfrequency current ripple on supercapacitors ageing through floating ageing tests.Microelectron Reliab 2013;53(9–11):1643–7.

[16] Maxwell technology, K2 series ultracapacitors datasheet, Document number:1015370.3.2009.

[17] Meyer, Method for the production of fuel gas, 4,936,931, 1990.


http://refhub.elsevier.com/S0026-2714(15)30014-7/rf0075













































study on specific effects of high frequency ripple...

Documents