**3. Results**

#### *3.1. Solvent and Electrolyte Characterization*

Three dioxolanes with different substituents in position 2 were synthesized performing ketalization reactions between lactic acid (LA) and two different aldehydes (formaldehyde R,R' = H, acetaldehyde R,R' = H,Me) or acetone (R,R' = Me). This panel of substances was chosen to correlate any performance differences to small structural variations. To indicate the different solvents and simplify the discussion, the abbreviation LA-R,R' will be adopted, where LA indicates the lactic acid fragment and R,R' explicit the substituents in position 2 (Figure 2).

**Figure 2.** DOLOs synthesis and investigated solvent. 5-methyl-1,3-dioxolan-4-one LA-H,H; 2,5 dimethyl-1,3-dioxolan-4-one LA-H,Me; 2,2,5-trimethyl-1,3-dioxolan-4-one LA-Me,Me.

The ketals synthesis is an equilibrium reaction, and therefore it was necessary to use a Dean-Stark trap to remove water from the reaction mixture and favor the products formation. A synthesis that implies formation of water as a by-product, such as the ketalization reaction, is preferable to other synthetic pathways as water does not contribute to the E-factor of the reaction. Nevertheless, in order to increase the sustainability of the whole process, water-removal technologies that do not involve the use of solvents, such as pervaporation, will be explored in the future [49].

The investigated compounds are potentially obtainable from renewable sources, as there are green industrial routes to lactic acid and the other reactants starting from bio-based feedstocks [50,51].

However, to consider these compounds as sustainable solvents, it is essential to evaluate their intrinsic safety (flammability, toxicological profile), and the environmental impact in case of accidental release (ecological profile). To assess their flammability, the flashpoint was estimated using the method reported by R.W. Prugh [52]. Table 1 reports the calculated flash point of the produced DOLOs, their boiling temperature at atmospheric pressure, and

the stoichiometric concentration in air (CST) expressed as a volume percentage used to estimate the flash points.


**Table 1.** DOLOs and commercial EDLC solvents flashpoints.

\* Experimental literature data.

DOLOs have an intermediate flash point between those of the two traditional EDLCs solvents (fpACN = 2 ◦C; fpPC = 132 ◦C) and are hence suitable for applications in energy storage devices. The complete evaluation of their ecotoxicological profile is beyond the aim of this study, but it is possible to make some considerations based on their chemical nature. Ketals are a class of substances stable in an alkaline environment but labile in an aqueous and acidic media. Therefore, it is possible to assume that in an atmospheric and physiological environment, these functional groups rapidly hydrolyze, returning the parent reagents [53]. Lactic acid, acetone, acetaldehyde, and formaldehyde are regularly included in the cellular metabolic pathways at physiological concentrations, and have a low persistence in the natural environment. As an example, the human body produces about 50 g of formaldehyde per day [54]; the half-life of blood plasma formaldehyde is 1.5 min [55] and that of atmospheric formaldehyde in daylight is 50 min [56]. Therefore, the investigated solvents are sustainable alternatives also from the point of view of their ecotoxicological profile.

A preliminary evaluation of their electric performance was carried out by measuring the ionic conductivity and the electrochemical stability window of 1M TEMABF4 solutions. The results are reported in Table 2.


**Table 2.** Electrolyte characterization.

Large variations of ionic conductivity were recorded according to the nature of the substituents in position 2. Indeed, LA-H,H showed a conductivity of 8.5 mScm<sup>−</sup>1, while in presence of additional methyl groups, the conductivity gradually dropped to 1.5 mScm−<sup>1</sup> (LA-H,Me) and 200 μScm−<sup>1</sup> (LA-Me,Me). Furthermore, LA-Me,Me and LA-H,Me demonstrated a lower solvent capacity because saturated solutions were obtained at concentrations below 1 M, which are therefore unsuitable for applications in SC. A possible application of LA-H,H in Li-ion based devices was also assessed, as 1M solutions of LiBF4 displayed a conductivity of 2.1 mScm<sup>−</sup>1.

In addition to an adequate conductivity, EDLC applications require that the electrolyte has a wide electrochemical stability window (ESW), that can be preliminary estimated by Linear Sweep Voltammetry (LSW). A large ESW (Figure S5 and Table S1) was obtained for the electrolyte based on LA-H,H (ΔV 4.50 V, cut-off current densities 1 mAcm−2), comparable to that recorded for the PC-based electrolyte (ΔV 4.55 V, cut-off current densities 1 mAcm<sup>−</sup>2).

Therefore, LA-H,H was selected to evaluate the performance in a symmetrical EDLC with activated carbon based electrodes using TEMABF4 as conventional conducting salt.

#### *3.2. EDLCs Characterization*

Cyclic voltammetries (CVs) at different scan rates were performed to evaluate, respectively, the impact of the investigated electrolytes on the maximum operating voltage (OV) and capacitance retention (CR) of EDLCs. For a homogenous comparative analysis, Figure 3 shows the data with the LA-H,H-based electrolyte, and those of an EDLC prepared with TEMABF4 1 M in PC.

**Figure 3.** Cyclic voltammetry of LA-H,H and PC based electrolytes EDLC. (**a**) CV of LA-H,H TEMABF4 1M at different scan rate: 5, 25, 50, 100, 200 mVs<sup>−</sup>1. (**b**) CV of PC TEMABF4 1M at different scan rate: 5, 25, 50, 100, 200 mVs<sup>−</sup>1. (**c**) Capacitance retention at different scan rate of EDLCs containing investigated electrolytes. (**d**) Coulombic efficiency at different cell voltage of EDLCs containing investigated electrolytes.

Initially CVs were recorded at 5 mVs−<sup>1</sup> from 0 to 1 V, and the final potential was gradually increased by 0.2 V in different cycles and described by a rectangular-like shape typical of SCs devices, which deviates from a perfect rectangle (typical of an ideal capacitor) due to the resistance parameters [46] (Figure S6). The screening was stopped at 2.6 V when a Coulombic Efficiency (CE) close to 95% was achieved, which was chosen as the minimum efficiency threshold. Figure 3d shows the CEs as a function of the applied potentials, and the trend is the same for the electrolytes. The decrease of the EC with the increase of the operating potential is due to a gradual contribution of parasitic and irreversible faradic reactions that interfere in the charging capacitance and are absent in discharge capacitance. At 2.6 V the LA-H,H-based EDLCs provided a CE of almost 94.4%, which was slightly lower than that recorded with the PC-electrolyte (95.4%).

After establishing the OV of 2.6 V for both electrolytes, CVs were performed from 0 to 2.6 V, ranging the scan rate from 5 to 200 mVs−<sup>1</sup> (Figure 3a,b) to evaluate their capacitance retention (CR). The increase of the scan rate strongly influences the response of the EDLCs during CVs; in fact, a marked distortion of the rectangular shape was recorded due to the increase of the ESR, while the decrease of the specific capacity is caused by a lower availability of the electrodes surface. Indeed, at high scan rates the formation of the electric double layer is limited by a lower accessibility of the electrolyte ions in the porous structure of the electrode [57]. As reported in Figure 3c, LA-H,H/TEMABF4 resulted in a greater performance than PC/TEMABF4. At 200 mVs−<sup>1</sup> a CR of 71% was achieved for LA-H,H/TEMABF4, while a CR of 62% were obtained with PC/TEMABF4. Based on these results, the increased performance of LA-H,H/TEMABF4 cannot be justified by its conductivity, as it was slightly lower than the PC-electrolyte. To understand this behavior, the solvent–salt and electrolyte-electrode interactions should be investigated in depth, however this aspect is beyond the scope of this paper and will be the subject of future investigations.

The storage properties and the internal resistance parameters of the investigated EDLCs were assessed through GC and EIS analysis (Figure 4).

**Figure 4.** Charge/discharge GC and EIS of LA-H,H and PC based electrolytes EDLCs. (**a**) GC of LA-H,H TEMABF4 1M at different current densities: 0.5, 1, 2, 5 Ag−1. (**b**) GC of PC TEMABF4 1M at different current densities: 0.5, 1, 2, 5 Ag<sup>−</sup>1. (**c**) EIS Nyquist plot of EDLCs containing investigated electrolytes for a frequency range from 500 kHz to 10 mHz. (**d**) Evolution of the imaginary part of the complex capacitance vs frequency of the same EDLCs.

The GCs were performed ranging the current density from 0.5 Ag−<sup>1</sup> up to 5 Ag−<sup>1</sup> to test the electrolytes in different conditions (Figure 4a,b). In all applied conditions the GC profile resulted in a symmetrical triangular shape, indicative of a good reversible and capacitive behavior. From the ohmic drop and discharge profile, the ESRGC and specific capacitance of each analysis was estimated, respectively. These results were used to calculate specific maximum power and energy.

The results obtained at low current density are summarized in Table 3.


**Table 3.** GC results at 0.5 Ag−<sup>1</sup> normalized with total mass of electrodes active materials.

An analysis of the overall performance discloses that LA-H,H is competitive with PC. A modest decrease of capacitance and specific energy was obtained with LA-H,H/TEMABF4 (entries 1 and 2), although the specific power dropped from 29.1 kWkg−<sup>1</sup> of PC (entry 1) to 22.5 kWkg−<sup>1</sup> of LA-H,H (entry 2) due to a higher ESRGC of the latter. Moreover, a slight decrease of CEGC was obtained from discharge/charge time ratio. At a high current density (5 Ag−1), the specific energy and power for LA-H,H/TEMABF4 reached 6 Whkg−<sup>1</sup> and 23 kWkg−1, while for PC/TEMABF4 the values were 14 Whkg−<sup>1</sup> and 29 kWkg−1. It is therefore evident that the solvent-salt interaction can significantly affect the performance of the electrolyte, especially the resistance parameters.

To investigate possible variations in resistance parameters, EIS measurements were performed by applying small perturbations (5 mVAC) and spanning the frequency range from 500 kHz to 10 mHz (Figure 4c). The Nyquist plot can be divided into high, medium, and low frequency parts in which the EDLC behavior transits from a completely resistive behavior to a completely capacitive one. This change is highlighted by the time constant τ<sup>0</sup> = 1/ν0, where ν<sup>0</sup> is the frequency with the highest imaginary part of the complex capacitance. From the analysis of these profiles, it was possible to separately determine the Equivalent Series Resistance ESREIS, the Equivalent Distributed Resistance (EDR) relative to the penetration of the ions into the electrode pores, and the bulk resistance of the electrolyte (Rbulk) related to electrolyte conductivity.

The resulting profiles of EISs experiments are typical of an EDLC device; in fact, at high frequencies there is a purely resistive behavior, and from the first intersection with the x-axis this can be defined as the ESR. Immediately afterwards, with slightly lower frequencies, a hemicycle shape of a mixed resistive/capacitive behavior begins to be visible due to the accumulation of charge at external electrodes surface and the relative resistance due to the charges transfer from bulk to the electrodes. The segment under the hemicycle has been defined as Rbulk due to the preponderant contribution of the electrolyte solution on this parameter. At intermediate frequencies the impedance profile tends to linearize with an almost 45◦ slope; this property is typical of a diffusive process, and considering the EDLC device is interpreted by the penetration of the ions in the porous structure of the electrodes. At low frequencies the electric double layer is able to structure itself completely, occupying the entire available electrode surface and saturating its internal volume. In this condition, an almost complete capacitive behavior is recorded, highlighted by a quasi-vertical profile. EDR was therefore defined as a segment of resistance between the intercepts with the x-axis by the linear fittings of the diffusive (medium frequencies) and capacitive (low frequencies) behavior.

The parameters obtained were normalized for the surface of the electrodes, and are summarized in Table 4.


**Table 4.** EIS results normalized for the electrode surface.

According to the previous data obtained from the GC analysis and conductivity evaluations, the resistance parameters obtained with the PC-based electrolyte (entry 1) were lower than those achieved with LA-H,H-based electrolytes (entry 2). Among the

investigated parameters, Rbulk and EDR were greater affected by the solvent nature of the electrolytes, while ESREIS was not particularly sensitive, resulting therefore more highly influenced by the electrode.

The EIS analysis was also used to determine the time constants of the EDLCs. Figure 4d shows the profile of the imaginary part of the complex capacitances as a function of the applied frequency used to determine the time constants. As expected, PC/TEMAB4 displayed a smaller time constant compared to that of LA-H,H/ TEMAB4, respectively 5 s and 10 s.

Finally, the stability over long-cycling of the electrolyte LA-H,H was evaluated by performing 5000 charge/discharge cycles at different current densities (1000 cycles for each current density). The evolution of capacitance, coulombic efficiency and capacitance retention over each cycle are reported in Figure 5.

**Figure 5.** (**a**) Evolution of specific capacitance, (**b**) coulombic efficiency and (**c**) capacitance retention of EDLCs containing LA-H,H/TEMABF4 at different current densities.

As shown in Figure 5a, initial loss of capacitance was observed for the first two groups of cycles, which then stabilized in subsequent cycles. Figure 5c shows the capacitance retention evaluated by the ratio between the specific capacitance of each cycle and the specific capacitance of the first cycle of each group of cycles. A minimum CR of 94% was obtained in the first thousand cycles at 0.5 Ag−1. At 1 Ag−<sup>1</sup> the efficiency range dropped to a minimum of 90%, while for the further cycles at 2, 5 and again at 0.5 Ag−<sup>1</sup> the final efficiencies were respectively 96.4%, 98.7% and 95.4%.

Furthermore, Figure 5b shows the coulombic efficiency over all cycling-groups, which was highly stable and close to 100% for all applied current densities, validating the stability of the electrolytes based on LA-H,H. To ensure a prompt comparison between the achieved results with other electrolyte categories, some EDLCs-electrolytes performances reported in literature are collected in Table S2.
