*3.1. Electrospun Separator and Electrolyte Selection*

Figure 1a,b report the SEM images of the electrospun Pu and CTA membranes, respectively. They feature interconnected fibers, randomly deposited, with a low number of defects. The Pu mat thickness was 55 μm and the mean fiber diameter was around 0.3 μm. The CTA mat thickness was 22 μm and the mean fiber diameter was around 0.6 μm. The fiber thickness of the two mats is in line with the value already reported for electrospun separators obtained with different polymers [33]. Furthermore, the PU and CTA mat thicknesses were adequate for an easy handling and assembly of the EDLCs. In addition to the difference in fiber diameter, the two polymers differ in terms of fiber diameter distribution, the CTA fibers being less homogeneous with a broader distribution.

Before the evaluation of the ionic conductivity response of the membranes, at first bulk conductivity of the electrolytes was measured. The values at different temperatures are reported in Table 2. The ionic conductivity of all the tested electrolytes grows with temperature. Among the considered electrolytes, the most conductive one is the EmimTFSI. Specifically, at 30 ◦C EmimTFSI features 12.6 mS cm−1, which is 5-fold higher than the conductivity of 0.5 m LiTFSI in TEGDME (2.05 mS cm<sup>−</sup>1) and PYR14TFSI (3.01 mS cm<sup>−</sup>1).

The separators of the EDLCs should be designed in order to achieve low ESR. This can be obtained by minimizing their hindrance to the ion flow during the charge/discharge, while guaranteeing the electronic separation of the two electrodes.

In order to evaluate the contribution of the investigated separators and electrolytes to ESR, EIS measurements were performed. The tests were carried out using cells with stainless steel blocking electrodes separated by the separator soaked with the electrolyte. EIS was carried out at constant interval of time (24 h) and at different temperature (30 ◦C, 40 ◦C and 60 ◦C) to check the chemical and electrochemical stability of the different membranes in the tested electrolytes.

**Figure 1.** SEM images of electrospun membrane of (**a**) pullulan and (**b**) cellulose triacetate with their molecular structures.

**Table 2.** Ionic conductivity of the tested electrolytes at different temperatures.


As an example, Figure 2 reports the Nyquist plots of the electrospun pullulan separator in EmimTFSI over time at the different tested temperatures. The Nyquist plots for all the combination of Pu and CTA membranes with the different electrolytes are reported in Figure S2.

**Figure 2.** Nyquist plot of Pullulan electrospun membrane in EmimTFSI electrolyte.

In Figure 2, the impedance spectra of the Pullulan membrane resemble a straight line. This response can be modelled with a resistance (R) in series with a constant phase element (Q), therefore the resulting impedance is given by the following equation:

$$Z = \mathbb{R} - 1/(\text{j} \text{ }\omega \text{ Q})^n \tag{6}$$

In Equation (6), R is the equivalent resistance of the separator soaked in the electrolyte and can be evaluated from the intercept with the real axis in the 150–300 kHz frequencies region. It includes the electronic resistance of the current collectors and the ionic resistance of the cell which reasonably dominates the response. When n = 1, the plot is a line parallel to the imaginary axis and Q represents the capacitive response of the cell. When n = 0.5, the plot is a line with a slope of 45◦ and Q corresponds to the Warburg element that is representative of diffusion-controlled processes.

Figure 2 shows that the temperature increase leads to the decrease of the resistance of the cell that is related to the increase of the electrolyte conductivity (cf. Table 2). In parallel, it is noticeable that the slope of the Nyquist plot decreases, unvealing that ion diffusion through the membrane becomes more sluggish.

This behaviour could be explained with the swelling of the membrane at the highest temperature that, in turn, brings about thickening of the fibres and narrowing of the inter-fibre voids. This might result in a more tortuous path for ion conduction.

Tables S1 and S2 and Figure 3a,b report the values of resistance of Pu and CTA membrane respectively, at different temperatures over time, in the different electrolytes. The values are in the same order of magnitude and span between ca. 2 and 5 ohm cm2. The first day at 30 ◦C, Pu features 2, 3 and 3.5 Ohm cm<sup>2</sup> when soaked with EmimTFSI, 0.5 m LiTFSI TEGDME and PYR14TFSI, respectively. CTA exhibits 2, 3 and 3.5 Ohm cm<sup>2</sup> with EmimTFSI, 0.5 m LiTFSI TEGDME and PYR14TFSI. Therefore, resistance values are similar for both membranes in the same electrolytes, with EmimTFSI accounting for the smallest values. A more straightforward comparison must consider the mat thickness of both separators and can be carried out referring to the effective resistivity (<sup>e</sup>ff) of the membrane-electrolyte system. The value of <sup>e</sup>ff can be obtained by Equation (7):

$$
\rho\_{\rm eff} = \mathbf{S} \times \mathbf{R} / \mathbf{L} \tag{7}
$$

where R is the resistance (in Ohm), L is the membrane thickness (cm), and S is the current collector area (cm2).

As commented above, Pu separator features a thickness of 55 μm that is almost 2.5 times larger than the CTA's that is 22 μm. Therefore, <sup>e</sup>ff of Pu at 30 ◦C in EmimTFSI results 450 Ohm cm and is almost half than CTA's (over 1000 Ohm cm). This can be related to the thinner fibres of the former membrane (0.3 μm) vs. the latter (0.6 μm). Thinner fibres provide a greater surface area and a greater density of free volume that can be exploited by ions to achieve higher conductivity. Noticeably, the resistance values of Pu at the different temperatures keep almost constant during time. At the contrary, those of CTA membrane gradually increase achieving 5 Ohm cm<sup>2</sup> at 60 ◦C, after 5 days, a value that doubles the Pu ones. Furthermore, after six day, the temperature was lowered to 30 ◦C. The resistance of Pu-EmimTFSI went back to its initial value while the CTA-EmimTFSI ones doubled (4 Ohm cm2). This indicates that the swelling process promoted by the increase of temperature is reversible for Pu but not for CTA. Overall, the data of Figure 2 suggest that Pu membrane is more stable than CTA.

In order to get further insight into the contribution of the separator to the ESR, the Mac Mullin number (NM) has been calculated for all the tested systems. Indeed, NM quantifies the increase of resistivity of the separator soaked in the electrolyte (<sup>e</sup>ff) with respect to the bulk resistivity of the electrolyte solution (-0), and it is calculated after Equation (8):

$$\mathcal{N}\_{\mathsf{M}} = \varrho\_{\text{eff}} / \varrho\_0 \tag{8}$$

where <sup>e</sup>ff has been evaluated by Equation (6) using the resistance values listed in Tables S3 and S4. In turn, -<sup>0</sup> is the reciprocal of the electrolyte conductivity (σ0) and is calculated by Equation (9):

$$
\varrho\_0 = 1/\sigma\_0 \tag{9}
$$

The NM values for the different separator/electrolyte combinations at the different temperatures are reported in Tables S5 and S6 and in Figure 3c,d as comparative histograms.

**Figure 3.** Resistance normalized by the plain area and MacMullin number of (**a**,**c**) Pullulan and (**b**,**d**) Cellulose triacetate electrospun separators in different tested electrolytes.

The values of the Pu are always smaller than those of the CTA in all the tested condition. For both membranes, in all the tested conditions, EmimTFSI holds the greater values of NM, while the smaller ones are exhibited by 0.5 m LiTFSi in TEGDME. The first day at 30 ◦C, Pu features NM of 5, 1 and 2 when soaked with EmimTFSI, 0.5 m LiTFSI TEGDME and Pyr14TFSI, respectively. For CTA, NM is 13, 3 and 5 with EmimTFSI, 0.5 m LiTFSI and PYR14TFSI. These trends indicate that EmimTFSI is the electrolyte that has a conductivity that is more affected by the presence of the membranes. In turn, this can be explained taking into account the protic behaviour of EmimTFSI. Indeed, unlike the other electrolytes, EmimTFSI features an acidic proton in alpha position in the imidazolium ring, that contributes to its bulk ionic conductivity. When EmimTFSI is in contact with the membranes this proton drives specific acid-base interactions that decrease its activity. Specifically, it can be claimed that hydrogen bond with the carboxyl functionalities of the membranes are formed (Figure 1).

For both separators soaked with EmimTFSI, NM increases with temperature. In case of Pu, it reaches a maximum of 11 on the day 3 at 60 ◦C. For CTA NM is 65 during the day 5 at the same

temperature. Once cooled at 30 ◦C (day 6), Pu-EmimTFSI's NM reversibly reduces to 4 that is even smaller than its initial value, in agreement with the resistance trend (Figure 3a). At the contrary CTA-EmimTFSI's NM does not recover its initial value and doubles (25).

To conclude this section, EmimTFSI-Pu featured a resistance considerably smaller than the one obtained with the other electrolytes. Pu ehibited a lower McMullin number than CTA along with a better thermal behaviour. Therefore, the pullulan based electrospun membrane and EmimTFSI were selected to assemble and test EDLCs as described in the next section below.

#### *3.2. Supercapacitor Testing*

The EDLCs featured the commercial high surface area carbon BP10 and the conductive additive Super C45. We already demonstrated the good binding properties of the pullulan: glycerol mixtures, that was therefore selected for the aqueous processing of the carbon composite electrodes [31].

The following sections report the electrochemical characterization of EDLCs assembled with 20% binder and low composite electrode mass loading (3.6–4.6 mg cm<sup>−</sup>2), referred as high binder low mass electrode (HBLME, Section 3.2.1), and with 10% binder and higher mass loading, labelled as low binder high mass electrode (LBHME, Section 3.2.2). The first composition was meant to verify the feasibility of the use of Pu binder and Pu membrane in the tested electrolyte while the second is meant to reach a formulation closer to that exploited commercial EDLCs. Section 3.2.3 compares the performances of HBLME and LBHME based EDLCs.

The electrochemical tests at first included EIS measurements of both the individual electrodes and of the full cell. These tests enable the evaluation of the EDLCs ESR that accounts for the contributions of (i) the contact resistance between composite material and current collector and (ii) the ionic resistance of the separator/electrolyte. Two electrodes cyclic voltammetry (CV) experiments have been carried out between 0 V and 3.2 V to evaluate the electrochemical stability and the capacitance of the EDLC as function of the scan rate. Galvanostatic (GCPL) charge/discharge measurements between 0 and 3.2 V (GLV) at different specific currents were subsequently performed to evaluate the specific energy and power. Finally, GCPL cycling has carried out at1Ag−<sup>1</sup> in order to evaluate the stability of the proposed EDLCs.
