High-energy density aqueous supercapacitors: The role of electrolyte pH and KI redox additive Open Access

Supercapacitors (SCs) represent a wide class of energy storage systems that, in general terms, can be charged/discharged quickly (in the second/sub-second timescale) over a large number of re-charge cycles, while retaining their initial performance.^1,2 They are mainly categorized in electrochemical double layer capacitors (EDLCs), which rely on electric double-layer formation via ion adsorption and swapping of co-ions for counter-ions at the electrode/electrolyte interface,^3,4 and pseudosupercapacitors, which mainly store electrical energy via faradaic electron charge-transfer by means of redox reactions, intercalation or electrosorption phenomena.^5–7 The major drawbacks of commercial EDLCs are their insufficient energy density (generally <15 W h kg⁻¹) and fast self-discharge (voltage losses between 5% and 60% over 2 weeks)^8,9 compared to those of batteries (energy densities between 30 and 270 W h kg⁻¹ and self-discharge rate typically lower than 10% per month).^10,11 Aqueous EDLCs, i.e., EDLCs based on aqueous electrolytes, have been also proposed to reduce the costs and potential hazards of traditional configurations, while exhibiting superior capacitance and rate capability, because of the smaller-size ions and higher ionic conductivity, respectively.^12,13 Nevertheless, the thermodynamic voltage required to decompose water is 1.23 V,^12,13 which limits the maximum voltage of aqueous EDLCs, and thus, their overall energy density, compared to traditional organic-electrolyte-based EDLCs.¹⁴ Potentially, several pseudocapacitive materials can be used to increase the energy density of aqueous SCs up to values comparable or even superior to those of batteries,¹⁵ and asymmetric SCs based on carbon and NiOOH electrodes (Saft Nickel Capacitors) have been also commercialized.¹⁶ However, Coulombic/voltage/energy efficiencies (CE/VE/EE) and cycle stability of pseudosupercapacitors are typically limited compared to traditional EDLCs,^6,7 hindering, so far, a massive market uptake. In fact, despite the enormous literature on pseudosupercapacitors remarking their excellent capacity (or effective capacitance) performances, the partial irreversibility of faradaic charge processes and the concomitant irreversible redox reactions limit the CE and the stability of the devices, while overpotentials of “pseudocapacitive reactions” limit their VE, and, thus, the EE at the power densities at which EDLCs typically operate. Such aspects also highlight the need for good practice for the unambiguous assessment of the real performance of these devices, as also discussed in recent literature.^17–19 To solve these issues, hybrid ion capacitors, including Li-ion capacitors, combine EDLCs electrodes with metal-ion intercalating electrodes to increase the energy density and limit the self-discharge of traditional SCs, without excessively sacrificing their rate capability.²⁰ Nevertheless, the rate capability and the cycle life of commercially available hybrid capacitors, i.e., Li-ion capacitors,²¹ depend on the Li⁺ diffusion dynamics in battery-type electrodes,²⁰ which must be designed with nanostructured materials to decrease Li⁺ diffusion length, introducing carbon coating and doping elements to maintain high electrical conductivity.²² Meanwhile, likely inspired by redox-flow batteries, redox-enhanced electrochemical capacitors have been also designed by adding soluble redox couples in the electrolyte, which becomes an active component for charge storage.^23,24 In fact, the redox species actively contribute to faradaic charge storage mechanisms, improving the energy density of traditional SCs.^23–25 Despite these advancements, SC manufacturers are still resilient in proposing redox-enhanced SC configurations, which often need costly ion-selective membranes to mitigate self-discharge phenomena (e.g., shuttling of mobile redox species)^25,26 or exhibit overpotential diffusion losses that limit the device EE.^23–25 Overall, it is still critical to overcome the classic energy density/power density (or cost) dichotomy in SCs,²⁷ and universal research efforts are on-going on this objective, while proposing cost-effective and sustainable device configurations,²⁸ including those based on sustainable and hazard-free water-based electrolytes.

In this work, we report a systematic study on aqueous EDLCs (based on a mixture of activated carbon and graphene as the active materials^27,29), screening acidic, neutral, and alkaline electrolytes, as well as the addition of a prototypical redox additive, i.e., KI, debunking the myth that aqueous SCs exhibit low cell voltage, and, thus, low energy densities. In agreement with previous literature, the use of moderately corrosive near-neutral electrolytes can enlarge the electrochemical stability window of carbon-based EDLCs,^30–32 even allowing the application of cheap metallic current collectors.^31,33,34 The high electrochemical stability window of carbon-based EDLCs based on near-neutral pH electrolytes is in-depth discussed, proving a maximum charging voltage (MCV) above 2 V in 6M LiNO₃, leading to an EDLC energy density even higher than 18 W h kg⁻¹ at a power density up to almost 7 kW kg⁻¹. The addition of KI as a low-cost redox additive is further investigated to improve the performance of the aqueous EDLCs using redox-enhanced configurations. By monitoring the positive and negative electrode potential, we point out the importance of balancing the masses (and so the capacities) of the electrodes to maximize the energy density of redox-enhanced electrochemical SCs, regardless of whether the configurations are symmetrical or asymmetrical.^35–38 Indeed, unbalanced SCs exhibit MCVs locked by the redox potentials of the redox couples at the positive electrode and by irreversible reaction potentials at the negative electrode. Capacity-balanced redox-enhanced SCs can reach a capacity as high as 38.2 mA h g⁻¹, corresponding to a ∼75.2% enhancement compared to the one measured without redox additive for a charge voltage of 2.1 V (21.8 mA h g⁻¹). However, the introduction of redox species introduces partially irreversible faradaic reactions, which strongly limit the MCV, decreasing the energy density to values inferior to those recorded for optimized EDLCs. In addition, the overpotentials of the faradaic reactions significantly decrease the EE of the redox-enhanced SCs compared to those measured for EDLCs. Overall, we provide a novel understanding on the design of affordable aqueous carbon-based SCs competing with commercial energy storage technology, including organic electrolyte-based EDLCs and batteries. At the same time, we remark the importance of using unambiguous metrics to correctly assess the performance of aqueous SCs, showing that basic EDLCs can outperform more advanced SC configurations in terms of energy density.

The active materials of the SC electrodes were obtained by mixing commercial activated carbon (AB-520, MTI Corp.) with single-/few-layer graphene (BeDimensional S.p.A.) produced by wet-jet milling exfoliation of graphite,^39,40 following the protocols reported in our previous studies (activated carbon:graphene weight ratio of 90:10).^27,29 The active materials were mixed with carbon black (>99.9%, Alfa Aesar) and polyolefin-grafted acrylic acid copolymer (aqueous binder, MTI Corp.) with a 85:10:5 weight ratio in water (solid content concentration of ∼0.3 g L⁻¹) until a homogeneous slurry was obtained. The SC electrodes were prepared by depositing the as-prepared slurry onto graphite paper (Papyex Flexible Graphite, Mersen) by doctor blading using a MSK-AFA-H200A coater (MTI Corp.). The resulting electrode films were dried in a vacuum oven (Binder, VD 53-UL), at 80 °C overnight to remove the water residuals. The SCs were fabricated by stacking two electrodes in a Swagelok-type cell based on 316 L stainless steel pistons, a polytetrafluoroethylene (PTFE) insulating body and fluoroelastomer sealing rings, using a glass fiber membrane (Whatman glass microfiber filter) as the separator. Various aqueous solutions with different pH, e.g., 3M H₂SO₄ (reagent grade, 95%–98% Sigma-Aldrich), 6M LiNO₃ (99.99%, Sigma-Aldrich), and 6M KOH (reagent grade, 90%, Sigma-Aldrich), were investigated as the (acidic, near-neutral, and alkaline) electrolytes. For the near-neutral 6M LiNO₃ electrolyte, KI was added as a redox additive at a concentration of 1M for the realization of redox-enhanced SCs. The mass loading of the EDLC electrodes was ∼5 mg cm⁻² to be compatible with practical requirements (the weight of active material in commercial EDLCs must account for more than 30% of the total mass of the packaged device).⁴¹ For redox-enhanced SCs, the mass loading of the electrodes was varied to balance the electrode capacities over their corresponding operating potential range, as discussed hereafter.

The electrochemical measurements were performed using a multi-channel potentiostat/galvanostat (VMP3, Biologic), equipped with a current booster unit (±10 A, Biologic). Linear sweep voltammetry (LSV) measurements were carried out on different substrates investigated as current collectors, including 316 L stainless steel (Sigma-Aldrich), graphite paper (Papyex Flexible Graphite, Mersen) and Ti (Sigma-Aldrich), to assess their electrochemical stability window. Cyclic voltammetry (CV) measurements were carried out on the SCs at various voltage scan rates, ranging from 10 to 400 mV s⁻¹. Galvanostatic charge–discharge (CD) measurements of the SCs were carried out at different current densities, ranging from 0.25 to 50 A g⁻¹. Three-electrode configuration measurements were also performed using positive (or negative) and negative (or positive) electrodes as the working and counter electrodes, respectively. The reference electrode was KCl-saturated Ag/AgCl for near-neutral and acidic electrolytes, and 1M NaOH-filled Hg/HgO electrode for alkaline electrolytes.

The specific capacity (A h g⁻¹) of the SCs was calculated from their galvanostatic CD curves as^18,19

where I (A) is the applied current, Δt (s) is the discharge time, and m (g) is the total mass of the electrodes. The $Im$ ratio is referred to as specific current. The specific capacity of a single electrode was estimated from the CD curves acquired with a three-electrode configuration cell using Eq. (1). In addition, the specific capacity of the SC (or the electrode) can be converted to an effective capacitance (C_eff) (F g⁻¹), i.e., the capacitance of an equivalent-capacity EDLCs (or capacitive electrode) by

in which ΔV (V) is the voltage window of the cell.

The discharge energy density (W h kg⁻¹), or discharge specific energy (hereafter referred to just as energy density), of the SCs was calculated using the following integral equation that considers the non-linearity of galvanostatic discharge characteristics:^18,19

where $Vt$ is the cell voltage (V) as a function of time (s), while t(V_max) and t(V_min = 0) are the moments in time corresponding to the maximum and minimum cell voltages (V_max and V_min), respectively. The charge energy density is calculated similarly to the discharge energy density, except that the integral is calculated over the galvanostatic charge characteristic [i.e., from t(V_min = 0) to t(V_max)].

The discharge power density (W kg⁻¹), or specific energy (hereafter referred to just as power density), of the SCs was calculated by^18,19

where t_D (s) is the discharge time extrapolated from the galvanostatic CD curve used for the calculation of the energy density.

The CE of the SCs was calculated by the ratio of the t_D and charge time (t_C) of the galvanostatic CD curve, i.e.,^18,19

while their EE, which can significantly differ from the CE in the presence of non-linear galvanostatic CD characteristics,¹⁸ was determined by the ratio between the discharge energy density and charge energy density, i.e.,^18,19

By the ratio between EE and CE, the VE can also be determined to quantify resistive and electrolyte diffusion-related losses.^18,19

Scheme 1 illustrates our approach to design high-voltage and high-energy density aqueous carbon-based SCs, consisting in (1) identification of electrolytes with different pH, possibly including KI as representative redox-additive, based on mechanistic considerations; (2) investigation of the electrochemical stability window of current collectors having different characteristics, including metallic and graphitic ones; (3) electrochemical characterization of carbon-based EDLCs by the progressive increase of V_max, until falling out of the electrochemical stability window of the devices, as monitored by a drop of CE; and (4) evaluation of KI redox additive on the electrochemical performance of the SCs, showing the importance of unequalizing the masses of the electrodes to balance their capacities and optimize the device energy density, comparing the results with those of reference EDLCs.

It is generally recognized that aqueous EDLCs show insufficient energy density compared to their organic counterpart because of their limited electrochemical stability window imposed by the thermodynamic voltage required to decompose water, which is 1.23 V.^12,13 Nevertheless, aqueous EDLCs based on stable electrode materials with a low catalytic activity toward water splitting reactions can be charged up to voltages higher than the thermodynamic one of water splitting. Carbonaceous materials, including activated carbon,^42,43 carbon black,³⁶ and graphitic ones,^35,44 have been demonstrated as examples of inert active materials for the development of high-voltage aqueous EDLCs.^35,36,42,43 The latter have been mainly demonstrated using near-neutral electrolytes, such as those based on alkali metal sulfates (Li₂SO₄,^32,45–49 Na₂SO₄^33,42,43,50 and K₂SO₄³⁶) and nitrates (NaNO₃⁵¹ and LiNO₃⁵²). The high electrochemical stability window of SCs based on near-neutral pH electrolytes has been associated with several factors. At the positive electrodes, the local oxidation of water leading to the oxygen evolution reaction (OER; H₂O → 1/2O₂ + 2H⁺ + 2e⁻) results in an acidification of the electrolyte that upshifts the equilibrium potential for the OER, which in turn widens the useable voltage window of the EDLC.⁵³ At the negative electrodes, the production of OH⁻ anions by the reduction of water leading to the hydrogen evolution reaction (HER; 2H₂O + 2e⁻ → H₂ + 2OH⁻) can result in a high local pH that downshifts the HER potential. In fact, as predicted by the Nernst equation, the shift of both HER and OER potentials is estimated to be pH × 0.059 V.^30,54 In addition, at negative potentials in moderate (e.g., near-neutral) pH media, mesoporous carbonaceous electrodes efficiently store the hydrogen generated by the water reduction step via electrosorption, impeding the completion of the HER and leading to reversible hydrogen electrosorption during CD cycling.^30,54 The rate of hydrogen adsorption/desorption reactions depends on the availability of free-water molecules near the electrode surface.⁵⁴ Accordingly, the formation of hydrated solid phases or highly viscous aqueous solution at high salt concentration may slow down the hydrogen desorption step, thus causing the progressive occurrence of irreversible HER. Meanwhile, a low salt concentration can limit the electrolyte conductivity, negatively affecting the EDLC rate capability. To mitigate the alleged dichotomy between hydrogen electrosorption rate and electrolyte conductivity, we selected 6M LiNO₃ aqueous solution as a near-neutral pH electrolyte for the realization of high-voltage carbon-based EDCLs. In fact, contrary to alkali metal sulfates that suffer from low solubility in water (e.g., 34.8 g per 100 ml at 20 °C for Li₂SO₄, leading to optimal ionic conductivity lower than 9 S m⁻¹ at a concentration of 2M),⁵⁵ alkali metal nitrates exhibit high water solubility, resulting in conductivity higher than 16 S m⁻¹ for LiNO₃ at a concentration higher than 4M.⁵⁶ Prospectively, other alkali metal nitrates can be used as critical raw material-free alternatives to LiNO_3. For example, NaNO₃ has a water solubility of ∼80 g per 100 ml at 20 °C, corresponding to maximum concentration higher than 9M and leading to satisfactory conductivity as high as 19 S m⁻¹ at a concentration of 5M. Beyond the 6M LiNO₃ electrolyte, 3M H₂SO₄ and 6M KOH aqueous solutions were investigated as representative acidic and alkaline electrolytes.¹² Finally, 1M KI + 6M LiNO₃ was also investigated as a representative near-neutral pH redox electrolyte for the formulation of redox-enhanced SCs, being KI a well-established redox-additive.^57–60 In general, the presence of soluble redox couples adds faradaic charge storage in addition to electric-double layer capacitance, thus allowing the utilization of the dead weight of the electrolyte as a active mass and the wasted pore volume in the electrodes as a reservoir for redox reaction products.²³ In addition, the use of halogen-based redox additives has been reported to be advantageous to limit the self-discharge of redox-enhanced SCs without requiring the use of expensive ion-selective membranes.²⁴ In fact, the negatively charged oxidation products (e.g., I₃⁻) are electrostatically held in the double layer of the positively charged carbonaceous electrode.²⁴ Moreover, the physical adsorption of the oxidized species within the surface of carbonaceous electrodes prevents ion cross diffusion.²⁴ Nevertheless, it is crucial to clarify the contribution of the faradaic charge storage at each electrode, as well as the voltage losses caused by limited redox reaction kinetics compared to that of electrochemical double layer formation. In particular, the use of a single (and viable) redox couple can be straightforward for the implementation of redox-enhanced SCs. However, the capacity of the positive and negative electrodes in symmetric SCs may result unbalanced because of the faradaic charge storage mainly associated with a single electrode. Therefore, depending on the concentration and utilization of the redox species, the capacities of the electrodes must be balanced via unequalization of the electrode masses to maximize the operating voltage of the device during CD cycles, thus maximizing the energy density. The EE metric must be also analyzed together with the CE to clarify how redox overpotentials can impact on the operating characteristics of the redox-enhanced SCs.¹⁸ 

Beyond the electrolyte properties, the current collector also affects the final performance of an SC, since its reactivity with the electrolyte can limit the device's electrochemical stability window. In particular, in aqueous electrolytes, the irreversible gas-evolving reactions can cause progressive delamination and/or disruption of the active material film.^61,62 In addition, the formation of a passive oxide film on the surface of a metallic current collector can deteriorate the electrical contact between the current collector and the active material film, to an extent that depends on the oxide film thickness and compactness.⁶² Figure 1 shows the LSV curves recorded for different substrates, including graphitic and metallic ones (i.e., 316 L stainless steel and Ti) in the investigated electrolytes. In 3M H₂SO₄ [Fig. 1(a)], Ti and graphite show similar onset potentials for the cathodic faradaic current densities. The latter is mainly attributed to the HER on both surfaces of graphite⁶³ and Ti.^64–66 Stainless steel shows a less negative onset potential for the HER, indicating its superior reactivity compared to both graphite and Ti. Stainless steel also shows less positive onset potential of anodic current densities, which can be attributed to both electrochemical corrosion and OER.^67,68 Graphite shows an anodic peak at 0.57 V vs. Ag/AgCl, which could be ascribed to the formation or even oxidation of oxygen functionalities.^69,70 Titanium shows negligible anodic current density, indicating poor activity toward parasitic oxygen-evolving and corrosion reactions. The outstanding corrosion resistance of Ti is associated with a thin (few-nm), amorphous, nonstoichiometric TiO₂ protective layer that is formed on the surface when exposed to an aerated environment.⁷¹ Thanks to its surface oxide layer, Ti also exhibits negligible anodic response in the other investigated electrolytes [Figs. 1(b)–1(d)]. Despite the limited reactivity, the semiconducting surface oxide may be associated with a high interfacial contact resistance with active material films, as shown from studies on other energy conversion devices (e.g., electrolysers⁷² and fuel cells⁷³). In 6M LiNO₃ [Fig. 1(b)], Ti shows the most negative onset potential for the cathodic current densities, even though the LSV data suggest a wide electrochemical stability window for both graphite (∼2 V) and stainless steel (>2 V). For stainless steel, the presence of an anodic peak at 1 V may be associated with a surface modification under anodic conditions, leading to nm-thick oxide surface layers.⁷⁴ A similar behavior of stainless steel was observed also in 6M KOH [Fig. 1(c)]. In this medium, the onset potentials of the cathodic current density for stainless steel and graphite are similar and less negative than the one of Ti.

Despite the superior inertness of Ti, its cost,⁷⁵ as well as the need for conductive coatings limiting its interfacial contact resistance with carbonaceous films,⁷⁶ may represent practical limits for its use as a current collector in SCs. Although the above observations are specifically oriented to the functional role of current collectors, it should be noticed that the electrochemical stability window of graphite can be considered as a limit case of the one of carbonaceous materials, for which a high specific surface area, structural and phase alterations of the graphitic sp² network, and oxygen functionalities, are expected to trigger irreversible faradaic redox reactions, in accordance with literature.^69,77 Based on this rationale, our data do not reveal a clear advantage in using graphite current collectors over the stainless steel ones, except for the case of 3M H₂SO₄, where the cathodic and anodic reactivity of stainless steel may shrink the electrochemical stability window of an SC. Nevertheless, the gas-evolving reactions on metallic substrates exhibited kinetics faster than the ones shown by graphite. Consequently, a pronounced gas evolution, as well as the passivation of the metallic surface, may cause the deterioration or insufficient electrical contact between the current collector and the active material films. To not jeopardize the number of SCs to characterize, and based on the above considerations, we selected graphite-coated stainless steel as the current collector for the realization of our full devices. By doing so, the hydrophobicity of graphite can impede the pronounced gas evolution in the proximity of the interface between the active material and current collectors, while avoiding the uncontrolled interfacial contact resistance between passivated metallic surfaces and active materials. Thanks to its high thermal conductivity (>100 W m⁻¹ K⁻¹ for typical non-pyrolytic 50 µm-thick graphite foils,⁷⁸ and even >1000 W m⁻¹ K⁻¹ for few µm-thick pyrolytic graphite foils⁷⁹), graphite can also provide a functional role for heat transfer during high-power operation of SCs. In addition, the metallic substrate lying below graphite can provide an optimal in-plane electrical conductivity (on the order of 10⁴ S cm⁻¹ for stainless steel) in practical devices, including pouch cells, which may be not reached by low-cost (i.e., non-pyrolytic) graphite (electrical resistivity on the order 10³ S cm⁻¹).⁷⁸ A similar approach has been reported in recent literature using graphite-coated Al current collectors for the fabrication of the aqueous SCs.³³ 

Figure 2 shows the CV curves measured for the carbon-based EDLCs based on the different aqueous electrolytes at a voltage scan rate of 400 mV s⁻¹, progressively increasing V_max until triggering irreversible faradaic reactions. The fifth CV scan is shown for each V_max, so that possible electrode conditioning occurring at the first cycles does not complicate the data interpretation. Regarding this aspect, a recent study showed that the potential of equal electrodes after assembling the EDLC can in fact differ from the potential of zero voltage (PZV), i.e., the equipotential of the electrodes when the EDLC is fully discharged (after being charged few times).³⁵ The preconditioning of the EDLCs by means of a few CV cycles can therefore involve irreversible faradaic reactions that progressively turn the behavior of EDLCs into non-faradaic one once the PZV is gradually regulated to fully exploit the capacitive potential window of both the EDLC electrodes.³⁵ Even if it is not the scope of our work, our data clearly indicate that the MCV of the EDLCs, here defined as the voltage limit of the EDLCs before the occurrence of irreversible faradaic reactions, is maximized to values well over 2 V in the near-neutral electrolyte (i.e., 6M LiNO₃), as expected from the above considerations on the selection of the electrolytes and the electrochemical stability window of the current collectors (see Fig. 1). Vice versa, the narrowest MCV (<1.8 V) was observed using the acidic electrolyte (3M H₂SO₄), which is consistent with the graphite reactivity in this condition [see Fig. 1(a)]. In 6M LiNO₃, the appearance of a distinguishable cathodic redox peak once the EDCL reaches a V_max beyond its 2.5 V suggests the partial reversibility of the faradaic redox reactions. As previously discussed, in moderate/near-neutral pH electrolytes, mesoporous carbonaceous electrodes can efficiently store the hydrogen generated by the water reduction step via electrosorption, impeding the completion of irreversible HER and leading to reversible hydrogen adsorption/desorption during CD cycling. Consequently, these effects are associated with the superior MCV of EDLCs based on near-neutral electrolytes compared to EDLCs based on acidic and alkaline electrolytes, where faradaic reactions are mainly associated with irreversible HER.^30,54

To carefully consider the non-linearity of the EDLCs caused by hydrogen electrosorption or other parasitic (redox) reactions, the performances of the EDLCs were evaluated in terms of specific capacity and efficiency metrics, which were calculated from galvanostatic CD curves at different specific currents (see calculation methods in the Experimental section).^18,19 Figures 3(a)–3(c) show the galvanostatic CD curves measured for the investigated EDLCs with different electrolytes at a specific current of 1 A g⁻¹, increasing the V_max until reaching values above their MCV. The latter condition is expressed by the loss of the nearly triangular shape of the CD curves and CE deterioration (i.e., t_D < t_C). Indeed, the appearance of a voltage plateau when approaching V_max is a clear indication of the occurrence of faradaic redox reactions that depart the EDLC from its ideal linear behavior.^18,19 Thus, these data confirm the CV results, evidencing that MCV is maximized for the EDLCs based on the 6M LiNO₃ electrolyte. Previous literature reported that the capacitances of positive and negative electrodes may be different, and the balancing of the electrode capacities by unequal active material mass loadings may be a strategy to maximize the overall specific capacity also for symmetric EDLCs.^36,38 Typically, the negative electrode exhibits a specific capacity higher than the positive electrode^36,38 (interestingly, as shown hereafter, we have not observed this behavior for EDLCs based on the near-neutral electrolyte). Consequently, in EDLCs, the positive electrode may reach the upper limit of its electrochemical stability window before the entire capacity of the negative potential is utilized, thus narrowing the device MCV.^36,38 Nevertheless, the self-regulation of the PZV during EDLC operation, as discussed above by referring to Ref. 35, may balance the operating potential window of the electrode to exploit the entire capacity of the electrodes. To verify such aspects, the electrode potentials were measured using a reference electrode to check the existence of a MCV-limiting electrode in the EDLCs. As shown in Figs. 3(d)–3(f), both electrodes reach a potential plateau while the EDLC voltage approaches the MCV, indicating that the electrode capacities are balanced. Interestingly, the PZV of the electrodes changes with varying V_max, especially once the V_max overpasses the MCV. These results confirm that the electrode ability to regulate the PZV enables a sort of smart self-optimization of the EDLC performances.³⁵ Moreover, considering the operating potential windows of the positive and negative electrodes and the equal active material mass loadings of the electrodes, these data indicate that, in the 6M LiNO₃ electrolyte, the positive electrode exhibits a C_eff higher than the one of the negative electrode, which is an opposite behavior compared to those observed in 3M H₂SO₄ and 6M KOH electrolytes. We speculate that the ratio of the C_eff of the electrodes strongly depends on the positioning of the PZV, which is difficult to predict in practice,³⁵ being dependent on the preconditioning of the EDLCs involving non-linear parasitic (redox) faradaic reactions.³⁵ 

Figures 4(a)–4(c) report the specific capacities measured for the EDLCs at various specific currents. At equal V_max (e.g., 1.5 or 1.7 V), the EDLC based on 6M KOH exhibits the highest capacity, because of the highest C_eff, e.g., 41.6 F g⁻¹ (or 166.4 F g⁻¹ for the single electrode) at 0.5 A g⁻¹. However, thanks to its superior MCV, the EDLC based on 6M LiNO₃ electrolyte can operate at higher V_max compared to the other EDLCs, reaching specific capacities even higher than 15 mA h g⁻¹ at V_max higher than 2 V for all the investigated specific currents. The highest specific capacity of 29.0 mA h g⁻¹, corresponding to a C_eff of 41.8 F g⁻¹ (or 167.2 F g⁻¹ for the single electrode), was therefore measured for the V_max of 2.5 V. Importantly, in 6M LiNO₃, C_eff significantly increases with the increase of V_max. For example, at 5 A g⁻¹, C_eff increases from 29.9 F g⁻¹ for V_max of 1.5 V up to the highest measured value of 41.8 F g⁻¹ for V_max of 2.5 V. These results indicate the presence of non-linear charge storage in our EDLCs, including hydrogen electrosorption,^30,54 as previously discussed. However, for a V_max of 2.5 V and specific currents lower than 5 A g⁻¹, it was not possible to acquire complete galvanostatic CD cycles because of the appearance of voltage plateaus below 2.5 V attributed to irreversible faradaic processes.

For the sake of clarity, because of these non-linear processes, the nomenclature of our EDLCs should be revised with a suitable terminology (e.g., pseudo-EDLCs or simply SCs). However, hereafter, we maintained the simple EDLC nomenclature so as not to complicate the readability of this work. Noteworthily, the specific capacities reached by our EDLC using 6M LiNO₃ electrolyte are comparable to those measured for the analog EDLCs based on organic:ionic liquid mixture electrolytes and operating up to V_max of 3.5 V.²⁷ Even though the V_max of our aqueous EDLCs is still inferior to the one reported for organic electrolytes or ionic liquids, the higher C_eff achieved using aqueous electrolytes compared to traditional electrolytes may counterbalance the MCV limitation. Beyond the specific capacity metric, the efficiency metrics are crucial parameters to consider for the assessment of EDLCs. More in detail, “practical EDLCs” must show a CE higher than 90% to exclude irreversible faradaic reactions, as well as concomitant self-discharging-related processes, e.g., ohmic leakage and charge redistribution through the diffusion of ions adsorbed at the surface of the electrodes.^80–82 In both 3M H₂SO₄ and 6M KOH-based devices, by decreasing the specific current toward the lower limit of 0.5 A g⁻¹, the CE shows a significant decay [Figs. 4(d) and 4(e)]. In 3M H₂SO₄-based EDLC, the CE drops to 48.4% at 1 A g⁻¹ for V_max of 1.7 V (which was not reached at 0.5 A g⁻¹ due to the presence of redox reactions). In 6M KOH-based EDLCs, the CE drops to 29.3% for V_max of 1.7 V [Fig. 4(f)]. On the contrary, in 6M LiNO₃-based EDLCs, the CE still approaches 90% for V_max as high as 2.1 V, proving an efficient operation at a V_max higher than 2 V. Beyond the CE, it is important to evaluate also the EE, which can substantially differ from the CE in the presence of non-linear charge storage mechanisms, as shown for pseudo-SCs, in which, despite a CE higher than 90%, the EE can be as low as 50% [Figs. 4(d)–4(f)].¹⁸ Clearly, at high specific currents (generally >10 A g⁻¹), the CE approaches 100% for all the EDLCs and the EE decreases with the increase of specific current as a consequence of resistive losses associated with the EDLC internal resistance and kinetic losses of possible reversible faradaic reactions. Vice versa, at low specific current, generally inferior to 10 A g⁻¹, the CE also contributes to the reduction of the EE. Despite this general trend, the EEs of the EDLC based on 6M LiNO₃ are higher than 80% for all the investigated specific currents for V_max inferior or equal to 1.9 V. In addition, at 1 A g⁻¹, EE still approaches 80% for V_max as high as 2.1 V, indicating a wide range of operating specific currents in which the EDLC can efficiently function with V_max beyond 2 V. Such distinctive features reflect the superior energy density of the EDLC based on 6M LiNO₃ compared to those of the EDLCs based on 3M H₂SO₄ and 6M KOH, as shown by the Ragone plot analysis [Figs. 4(g)–4(i)]. In particular, for V_max of 1.9 V, the EDLs based on 6M LiNO₃ exhibits energy densities higher than W h kg⁻¹ at power densities up to almost 7 kW kg⁻¹. For V_max of 2.1 V, the EDLC exhibits energy densities higher than W h kg⁻¹ at power densities up to almost 7 kW kg⁻¹. By further increasing the V_max up to 2.5 V, the EDLC can exhibit an energy density up to 27.7 W h kg⁻¹ at the power density of ∼26 kW kg⁻¹, while still preserving an EE of 70.0%.

Even though our metrics’ analysis indicated the possibility to realize aqueous high-voltage (>2 V) EDLCs with high energy density and excellent rate capability, their ability to withstand a great number of re-charge cycles without losing performance must be also proved. To understand the relation between the cycling stability of the EDCLs and their V_max, thousands of galvanostatic CD cycles were performed by progressively increasing V_max by 0.2 V, acquiring 1000 galvanostatic CD cycles for each V_max (Fig. 5). More in detail, V_max was varied between 1.3 and 1.9 V for both the EDLCs based on 3M H₂SO₄ and 6M KOH electrolytes, while the EDLC based on 6M LiNO₃ was tested for V_max ranging from 1.5 to 2.5 V. The galvanostatic CD curves were acquired at 1 A g⁻¹ for the 6M LiNO₃-based EDLC, while the specific current was increased to 2.5 A g⁻¹ for the other EDLCs, since they could not reach the investigated V_max when tested at 1 A g⁻¹, indicating an intrinsic operational limit of such devices. Both the EDLCs based on 3M H₂SO₄ and 6M KOH electrolytes showed a pronounced specific capacity degradation over galvanostatic CD cycling for V_max of 1.9 V [Figs. 5(a)–5(c)]. In addition, the CE of the devices was inferior to 90% for all the V_max ≥ 1.5 V. On the contrary, the 6M LiNO₃-based EDLC evidenced a stable specific capacity up to V_max of 2.1 V, with CE over 97% (≥99% for V_max up to 1.9 V) [Fig. 5(b)]. For V_max of 2.3 V, the CE progressively stabilized over 93%, but the specific capacity progressively decayed from 26.1 to 18.3 mA h g⁻¹. Because of possible modifications of the electrodes, the EDLC was then able to complete galvanostatic CD cycles at 1 A g⁻¹ even for a V_max of 2.5 V, which was not possible in fresh EDLCs [in accordance with data reported in Figs. 4(b) and 4(e)]. Since the performances recorded in the EDLC after being cycled with V_max of 2.3 V were already compromised, the specific capacities measured at 1 A g⁻¹ for V_max of 2.5 V (26.7 mA h g⁻¹ in the first cycle at such V_max) were inferior to that measured at 5 A g⁻¹ for the fresh EDLC (29.0 mA h g⁻¹). In addition, for a V_max of 2.5 V, the specific capacity degradation became even more pronounced compared to that for a V_max of 2.3 V. Overall, these data indicated that the EDLCs based on the 6M LiNO₃ electrolyte cannot safely operate at V_max higher than 2.1 V, which however is a satisfactory condition to provide an energy density higher than 18 W h kg⁻¹ at a power density up to almost 7 kW kg⁻¹ [Fig. 4(g)].

Once assessed the advantages of using near-neutral electrolytes for our aqueous carbon-based EDLCs, the addition of KI redox additive in the 6M LiNO₃ was subsequently evaluated, using 1M KI + 6M LiNO₃ as the redox electrolyte. As previously discussed, the use of soluble redox couples adds faradaic charge storage, providing extra capacity and maximizing the energy density of EDLCs.²³ In this context, Ref. 24 represents a seminal study on the key-rules for designing redox electrolytes for high-performance aqueous redox-enhanced SCs, identifying optimal redox couples for negative and positive electrodes, respectively. Importantly, Ref. 24 indicates that the use of halogen-based redox additives for the positive electrodes limits the self-discharge of redox-enhanced SCs without requiring the use of expensive ion-selective membranes.²⁴ Even if the use of two redox couples is favourable to maximizing the cell voltage, and thus the energy density of the redox-enhanced SCs, the second redox additive must be compatible with the first one, and it may increase the device costs. Therefore, redox-enhanced SCs based on a single redox additive, such as KI, are being widely studied.^57–60 Nevertheless, in the presence of a single redox additive, it is not trivial to predict the contribution of the faradaic charge storage of each electrode, as it depends on the PZV of the electrodes. In addition, if the faradaic charge storage mainly regards one electrode, the electrode capacities that were balanced in the previous EDLCs will not be anymore balanced in the resulting redox-enhanced SCs. Even more, the electrode with the smallest capacity may reach a potential outside its electrochemical stability window, limiting the MCV of the redox-enhanced SCs. Finally, the redox potentials of the redox couples strongly influence the operating voltage of the device during charging and discharging, affecting the device energy–power density characteristics.²⁴ To clarify the presence of these problems, we firstly investigated the redox-enhanced SCs based on electrodes with equal active material mass loadings, as we did in the previous section for the EDLCs. As shown in Fig. 6(a), the CV analysis of the redox-enhanced SC reveals the presence of multiple redox reactions that can be attributed to iodine/iodide, as well as iodide/iodate, redox pairs, according to the following reactions:^59,83–85 2I⁻_(aq) → I_2(s) + 2e⁻; I_2(s) + I⁻_(aq) ⇌ I₃⁻_(aq); I_2(s) + I₃⁻_(aq) ⇌ I₅⁻_(aq); I_2(s) + 6H₂0 → 2IO₃⁻ + 12H⁺ + 10e⁻, in which the subscripts (s) and (aq) indicate solid and soluble aqueous phases, respectively. The direct electrochemical generation of I₂ and I₃⁻ by 2I₃⁻_(aq) → 3I_2(s) +2e⁻ and 3I⁻_(aq) → I₃⁻_(aq) + 2e⁻, respectively, are likely excluded because of the concomitant precipitation of I₂ for a high concentration of I₃⁻, in accordance with I_2(s) + I⁻_(aq) ⇌ I₃⁻_(aq).⁸⁵ Nevertheless, the high anodic current densities above 1.2 V indicate partial irreversibility of the faradaic reactions, which may be cause of unsatisfactory CE, as shown hereafter. Since the decrease of the local pH is expected by 6H₂O → 2IO₃⁻ + 12H⁺ + 10e⁻, OER should play a marginal role in the irreversibility of the reactions at the positive anode. However, the imbalance of the electrodes’ capacities with the addition of a redox additive may have a strong effect on the MCV of the device. In fact, as shown by the CD curves reported in Fig. 6(b), which show a V_max of 1.5 V, the negative electrode reaches negative potentials outside its electrochemical stability window, triggering an irreversible faradaic reaction (i.e., HER). In addition, the potential of the positive electrode, blocked to values inferior to those of the EDLCs [see Fig. 3(e)], reflects a high low-voltage (<0.15 V) capacity (>60 mA h g⁻¹) that marginally contributes to the energy density of the device, in agreement with Eq. (3). Figure 6(c) shows that the specific capacity of the redox-enhanced SC is significantly enhanced compared to the one of the reference 6M LiNO₃-based EDLC, achieving values as high as 43.1 mA h g⁻¹ at 1 A g⁻¹. However, the redox-enhanced SC shows a drastic decay of the CE with decreasing the specific current, showing values lower than 90% for a specific current equal or inferior to 10 A g⁻¹ (CE = 57.6% at 1 A g⁻¹). In addition, the overpotential of the faradaic redox reaction causes voltage losses that, compared to the EDLC case, strongly lower the EE to values inferior to 61.3% for all the investigated specific currents.

To solve the issues of the redox-enhanced SCs with unbalanced electrode capacities (unbalanced device), we roughly estimated the C_eff of the positive and negative electrodes by CV measurements [Fig. 7(a)], limiting one side of the electrode potential windows to values nearby the PZV. Noteworthy, the latter is not strictly predictable, since it depends on V_max and on the device preconditioning, as previously discussed. Despite inevitable methodology limits, the data reveal that the C_eff of the positive electrode is ∼2.6 times the one of the negative electrode. Based on this analysis, we reduced the active material mass loading of the positive electrode by a similar factor, obtaining an electrode capacity-balanced redox-enhanced SC (balanced device). As shown from its CV curve in [Fig. 7(b)], the absolute intensity of the specific current at higher voltages increases compared to the unbalanced device. However, the issues related to the irreversibility of the faradaic reactions above ∼1.2 V are not solved. By referring to mechanistic working principles,⁸⁵ while solid I₂ is stored in the nanoporous structure of the carbon-based electrodes, mobile I₃⁻ and I₅⁻ are chemically formed. Owing to their negative charge, a certain fraction of iodides will remain adsorbed in the nanopores of the (positively polarized) positive electrode. However, beyond a certain limit, it is reasonable that an iodide fraction will diffuse toward the negative electrode, where they are reduced to I⁻ via the shuttling mechanism that cause the CE losses, as well as a self-discharge. Even if it is not the scope of the current work, the presence of inert anions, such as NO₃⁻ in our case, may increase the fraction of stored I₂, limiting iodide redox shuttle formation.⁸⁵ However, because of the equilibrium constant of I_2(s) + I⁻_(aq) ⇌ I₃⁻_(aq), finding the optimal redox additive concentration would be needed for avoiding the local concentration of I₃⁻ to be orders of magnitudes higher than the I₂ concentration at all times, as typically reported in the literature.^86,87 The control of the mobility of iodide species via engineering the nanoporous structure of the carbon-based electrodes may also represent a further strategy to mitigate CE losses.⁸⁵ As shown in Fig. 7(c), even if the unequalization of the electrode masses does not solve the irreversibility issues, the strategy slightly increases the capacity of the device compared to the unbalanced case. The balanced device was also tested decreasing the V_max to 1.1 and 1.3 V, enabling the operation in a potential window with reduced irreversibility. The data are compared also to those obtained for 6M LiNO₃-based EDLC, indicating that the increase of the specific capacity is mainly observed at low specific currents, depending on the V_max. Although these results prove that the specific capacity of redox-enhanced SCs can substantially improve the one of EDLCs, the Ragone plot analysis shows that the “simple EDLC” exhibits the highest energy density performance [Fig. 7(d)]. The origin of this behavior is attributable to its superior MCV, leading to a “high-voltage capacity” that mostly contribute to the device's energy density. Finally, Fig. 7(e) reports the efficiency metrics measured for the optimized redox-enhanced SCs, showing characteristics similar to those measured for the unbalanced device. Overall, our data show that the use of redox additive can increase the specific capacity performance, but implications in the MCV and self-discharge mechanism negatively affect both energy density and efficiency performances. Even if KI-based redox enhanced SCs are continuously reported in the literature,^57–60 our results evidence that they could not offer practical advantages compared to well-designed EDLCs. Prospectively, the use of a second redox additive with faradaic charge storage mainly associated with the negative electrodes may be a strategy to design effective redox-enhanced SCs with performances superior to those of EDLCs.²⁴ 

In summary, we reported an extended characterization of aqueous SCs, screening acidic, neutral and alkaline electrolytes, as well as the addition KI as a prototypical redox additive, performing both two- and three-electrode configuration measurements. The assessment of the device performances was carried out using proper metrics. The latter include specific capacity, energy and power densities, and EE. More in detail, energy and power densities and energy efficiency were calculated through methods based on the integration of the galvanostatic CD curves, avoiding the simplification of the device charging/discharging behavior to a linear one. Indeed, even in standard symmetric aqueous EDCLs, the occurrence of faradaic charge storage processes, such as hydrogen electrosorption, cannot be correctly analyzed by assuming a standard capacitive (linear) charging/discharging behavior. The rationale leading to the choice of electrolytes has been widely motivated by referring to mechanistic principles reported in the literature for aqueous SCs, showing the superior performance expressed by devices based on near-neutral electrolytes. By identifying proper substrate configurations, we experimentally proved that aqueous EDLCs based on near-neutral electrolytes (namely 6M LiNO₃) can reach an MCV superior to 2 V, enabling energy densities higher than 18 W h kg⁻¹ (comparable to/approaching those of lead acid and Ni–Cd batteries) at power density up to almost 7 kW kg⁻¹ (significantly superior to those of competing battery technologies). The introduction of redox additives can significantly increase the capacity of the SCs. However, even by optimizing the redox-enhanced SC performance by balancing the capacities of the electrodes via mass unequalization, both the MCV and the EE of redox-enhanced SCs significantly decrease compared to the simple EDLCs, mainly because of partially irreversible faradaic redox reactions and overpotentials of kinetically limited redox reactions, respectively. While debunking the myth that aqueous SCs exhibit low energy density, our study also remarks the importance of using non-misleading performance metrics to adequately assess aqueous SCs, e.g., redox-enhanced SCs, reporting a reliable set of device characteristics.

This project received funding from the European Union’s Horizon 2020 research and innovation program under Grant Agreement No. 881603-GrapheneCore3, the European Union’s Horizon 2020 Research and Innovation Program under the Marie Skłodowska-Curie Grant Agreement No. 813036, and the European Union’s SENSIBAT project under Grant Agreement No. 957273. We thank the Electron Microscopy and Material Characterization facilities—Istituto Italiano di Tecnologia—for support in SEM/TEM and XRD data acquisition, respectively.

The authors have no conflicts to disclose.

Matilde Eredia: Conceptualization (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Vittorio Pellegrini: Resources (equal); Supervision (equal). Francesco Bonaccorso: Conceptualization (equal); Funding acquisition (equal); Supervision (equal); Writing – review & editing (equal). Sebastiano Bellani: Conceptualization (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal). Marilena I. Zappia: Investigation (equal); Methodology (equal). Luca Gabatel: Investigation (equal); Methodology (equal). Valerio Galli: Methodology (equal). Ahmad Bagheri: Investigation (equal); Methodology (equal). Hossein Beydaghi: Investigation (equal); Methodology (equal). Gabriele Bianca: Investigation (equal); Methodology (equal). Irene Conticello: Investigation (equal); Methodology (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.