3.2.3. Cycling Stability, Energy and Power of HBLME- and LBHME-EDLCs

Figure 6a reports the trends of the specific capacitance of the two EDLCs over cycling at 1 A g<sup>−</sup>1. The values are normalized by the value of the specific capacitance of the first cycle. Both devices show a good stability with capacitance retention of 90% at the 2000th cycle. This result demonstrates the feasibility of the use of pullulan as alternative separator and binder for green supercapacitors.

Note that the cycling stability of LBHME-EDLC at low binder content was further evaluated even over 5000 cycles (Figure S3). The test indicated that also over prolonged cycling, a capacitance retention of 77% can be obtained.

The specific energy and power values of the two EDLCs, calculated through Equations (4) and (5) are compared in the Ragone plot reported in Figure 6b. Both devices deliver the maximum specific energy at the lowest current, and the maximum power is delivered at the highest current.

Indeed, at 0.5 A g−1, the specific energy is 19.6 Wh kg−<sup>1</sup> and 7.2 Wh kg−<sup>1</sup> for HBLME-EDLC and LBHME-EDLC, respectively. At4Ag−1, the specific power is 4.7 kW kg−<sup>1</sup> for the HBLME EDLC and 3.8 kW kg−<sup>1</sup> for the LBHME-EDLC. These values are reported in Table 3. If energy and power are normalized by the electrode area, they become 182 μWh cm−<sup>2</sup> and 42.8 mW cm−<sup>2</sup> for the HBLME-EDLC, and 99.4 μWh cm−<sup>2</sup> and 51 mW cm−<sup>2</sup> for the LBHME-EDLC (Table 3). These results clearly demonstrate that increasing electrode thickness is detrimental for energy and power performance. Indeed, specific energy of LBHME-EDLC is lower than HBLME-EDLC and this is mainly related to an inefficient exploitation of the electrode carbon surface. This is highlighted by the comparison of the area capacitance of HBLME-EDLC (148 mF cm<sup>−</sup>2) and LBHME-EDLC (85.5 mF cm−2). Therefore, in LBHME-EDLC, the increase of the electrode mass is not enough to offset such specific energy decrease, and the areal energy density keeps lower than that of HBME-EDLC. However, power performance of the two EDLCs are comparable, suggesting that the decrease of the binder content has a positive effect.

**Figure 6.** EDLC comparison by galvanostatic tests. (**a**) Trend of the capacitance percentage normalized by the value at first cycle the as function of the cycle number (at 1 A g<sup>−</sup>1, cell voltage cut-off: 0 V–3.2 V) and (**b**) Ragone plots of HBLME-EDLC and LBHME-EDLC.

### **4. Discussion**

Today many efforts are being devoted to increasing the specific energy of supercapacitors by different strategies. Among them promises are held by the use of ionic liquids, thick electrodes and pseudocapacitive active materials [2,6,12]. Ionic liquids enable high practical voltage (>3 V) and therefore energy density, but their major drawbacks are the greater ESR respect to commercial electrolytes and high cost. The use of thick electrodes (>10 mg cm<sup>−</sup>2) may seem the simpler solution, but the achieving high performance ith thick electrodes is still an unsolved problem [10,11]. Indeed, thick electrodes suffer of poor electronic and ionic connection between the particles, and only the external portion of the electrodes take part in the charge/discharge processes.

The increasing market for supercapacitors requires that sustainable manufacturing processes and materials are exploited to manufacture green supercapacitors. Aqueous processable bio-derived polymers represent a valuable alternative to today's fluorinated separators and binders. Furthermore, we have already demonstrated that the smart combination of a water processable binder and separator (like pullulan) and a hydrophobic ionic liquid electrolyte (EmimTFSI) enables an easy recovery of the expensive ionic liquid [31]. Indeed, the pullulan-IL-based EDLC can be readily separated into each of its components by immersion in water. After these very interesting but preliminary results, in this

paper we have carried out a study to get further insight into the impact of the use of natural polymer in supercapacitors performance. We have investigated the use of cellulose, which is the most widely studied bio-based polymer for green supercapacitors, and pullulan, that we have proposed for the first time in [31]. CTA and Pu separators were processed as self-standing mats by electrospinning. Their permeability to different electrolytes (EmimTFSI, 0.5 m LiTFSI TEGDME, PYR14TFSI), which is critical in the formation of the electrical double layer, has been evaluated by EIS and quantified referring to the Mac Mullin number. This characterization has been done at different temperatures over one week, in order to get insight on the thermal and chemical stability of the tested bio-polymer in the selected electrolytes. Both membranes at 30 ◦C featured the lower resistance (Pu 2 Ohm cm2 CTA 2 Ohm cm2) when soaked with EmimTFSI. However, given that the thickness of Pu (55 μm) was higher than that of CTA (22 μm), the resistivity of the Pu-EmimTFSI system was considerably smaller (0.407 kOhm cm vs. 1.041 kOhm cm). Furthermore, Pu-EmimTFSI exhibited a better thermal stability respect to the CTA-EmimTFSI. Notably, the NM for the Pu is always smaller than that of CTA, in particular, in EmimTFSI values of 5 and 12 were found, respectively. Overall, this study highlighted the presence of different and specific interactions between the tested electrolytes and the membranes that affect the ionic permeability and stability. It also indicated Pu-EmimTFSI as the best system. Indeed, Pu-EmimTFSI was the combination capable to minimize the ESR and avoid performance degradation due to temperature changes.

On the basis of these results, Pu was selected as separator and binder for EmimTFSI-based EDLCs. The big challenge we faced in this study was to reach high electrode mass loading at low binder content.

Two EDLCs have been assembled featuring two different formulations, one with lower mass loading and high binder content (HBLME) and a second one with higher mass loading and lower binder content (LBHME). These have been characterized electrochemically to evaluate how binder decrease and mass loading increase affect performance. In particular, these devices have been characterized at first by EIS. Analysis of the Nyquist plots highlighted an increase of the ESR moving from the HBLME-EDLC to the LBHME-EDLC, (5.9 Ohm cm<sup>2</sup> vs. 7.6 Ohm cm2). This trend was mainly related to the increase of the ionic and electronic contact resistances between carbon particles with the increase of electrode mass loading (1.5-fold from HBLME-EDLC to LBHME-EDLC).

For both devices' CVs have shown the absence of faradaic parasitic reactions within the cell voltage range 0 to 3.2 V. This wide range is feasible thanks to the good electrochemical stability of Pu-EmimTFSI. Specific capacitances have been calculated for both EDLCs. The highest specific capacitance was featured at the lowest scan rate of 5 mV s−<sup>1</sup> for both devices and was 18 F g−<sup>1</sup> and 14 F g−<sup>1</sup> for HBLME-EDLC and LBHME-EDLC, respectively. Both supercapacitors featured a good capacitance retention with the increase of the scan rate that however was higher for HBLME-EDLC (22%) than LBHME-EDLC (50%).

GLV cycling with high coulombic efficiency (higher than 98%) was demonstrated for both devices at the high cell voltage of 3.2 V. Noticeably, the EDLC featured a very good cycling behaviour demonstrated over more than 2000 cycles even at low binder content, confirming the stability of the Pu-polymer in EmimTFSI.

The high cell voltage and good specific capacitance provided specific energy of 19.6 Wh kg−<sup>1</sup> and 7.2 Wh kg−<sup>1</sup> at 0.5 A g−<sup>1</sup> that well compare with those of EDLCs featuring the same electrolyte and active carbon but employing a fluorinated binder and fiber glass separators [16]. The highest specific power was 4.6 kW kg−<sup>1</sup> and 3.7 kW kg−<sup>1</sup> at 4–5 A g−<sup>1</sup> respectively for HBLME-EDLC and for the LBHME-EDLC

The gravimetric performance of the HBLME-EDLC is superior respect to that of the LBHME-EDLC. This is due to the lower specific capacitance of the latter vs. the former. In turn, this is due to an inefficient ionic electronic contact between electrode carbon particles that leads to a partial exploitation of the electrodes surface.
