Fe- and Mg-modified binder-free Cu₂O thin films for solid-state symmetric supercapacitors: composition-dependent structure-electrochemical property relationships

Abstract

Binder-free cuprous oxide (Cu₂O) thin films were grown directly on copper foil by successive ionic layer adsorption and reaction (SILAR) using nominal Fe or Mg additions of 1, 3 and 5 at% relative to Cu. The films are denoted by modifier and nominal loading as Fe1, Fe3 and Fe5, and Mg1, Mg3 and Mg5, with the unmodified film written as pristine Cu₂O. Composition-matched electrodes were assembled into symmetric devices with a poly(vinylidene fluoride) (PVDF)-based LiCl/KOH gel electrolyte. X-ray diffraction shows a cuprite-dominant Cu₂O phase throughout both series, together with a composition-dependent shift of the (111) reflection and a corresponding change in coherent-domain size. Field-emission scanning electron microscopy shows compact granular Fe-modified films, whereas Mg modification produces a stronger evolution of particle size and surface architecture, culminating in an open agglomerated texture at Mg5. At 0.5 mA cm⁻², Fe1 and Mg1 deliver device areal capacitances of 129 and 80 mF cm⁻² against 12 mF cm⁻² for pristine Cu₂O, with fitted charge-transfer resistances of 759 and 18.5 Ω against (3.9 ± 0.7) × 10⁴ Ω for the unmodified film. The lowest charge-transfer resistances of the series belong to Mg1 and Fe5 rather than to the two highest-capacitance devices, and the Warburg coefficient falls by four orders of magnitude on modification. After 10 000 cycles all seven devices retain 95–99% of their capacitance, while coulombic efficiency remains strongly composition-dependent, from 61% for Mg1 to 97% for Fe5 and Mg3. Scan-rate current deconvolution assigns 59–94% of the stored charge to the surface-controlled term. Low-level Fe and Mg modification therefore tunes Cu₂O through distinct, non-monotonic combinations of interfacial kinetics, ion transport, redox accessibility and film morphology, and no single descriptor ranks the seven compositions in the same order.

Fe and Mg tune binder-free SILAR-grown Cu₂O in different ways: 1 at% Fe stores the most charge (129 mF cm⁻²) while 1 at% Mg gives the fastest interfacial kinetics (18.5 Ω), so capacitance and kinetics optimise at different compositions.

1. Introduction

Electrochemical capacitors bridge the power-energy gap between dielectric capacitors and batteries. Their value lies in rapid charge acceptance, high power delivery and extended cycle life, while the attainable energy is determined by the charge stored per unit mass, area or volume and by the usable cell voltage.^1–4 Charge storage can originate from electrostatic ion adsorption, rapid surface or near-surface redox reactions, or more strongly diffusion-limited processes. Distinguishing these contributions is particularly important for transition-metal oxides because similar voltammetric profiles can represent markedly different kinetic regimes.^4–6

Cu₂O is an attractive p-type oxide for low-cost energy-storage electrodes. Copper is abundant, the cuprite phase can be formed under mild chemical conditions, and the active oxide can be grown directly on a copper current collector. Direct growth removes polymer binders and conductive additives from the electrode-collector interface, shortens the electron pathway and preserves the geometric footprint of the device; related ordered oxides, oxide/graphene hybrids and conducting-polymer architectures exploit the same transport and surface-access principles.^7–12 Cu₂O-containing powders, microspheres, directly grown CuO/Cu₂O/Cu architectures and Cu₂O-based hybrid structures have consequently been evaluated for capacitive energy storage.^13–16 Performance varies widely with morphology, loading, electrolyte, voltage window and cell configuration, making device-level areal metrics especially useful for thin-film electrodes. Recent studies have also demonstrated promising strategies for improving the energy-storage performance and energy density of advanced supercapacitor and hybrid capacitor systems through rational electrode design, compositional optimization and improved charge-storage mechanisms.^17–19 Within this landscape, compositional tuning of a single, well-defined host oxide offers a controlled way to isolate the effect of the modifying cation from confounding changes in electrode architecture or formulation. In this context, the development of binder-free transition-metal-oxide electrodes with compositionally tunable structural and electrochemical properties remains an attractive approach for high-performance energy-storage devices. To our knowledge, no previous study has directly compared Fe- and Mg-modified Cu₂O prepared by the same binder-free SILAR route and evaluated as complete solid-state symmetric devices rather than three-electrode half-cells; the present work addresses this gap. A central finding of this comparison is that the composition which maximises stored charge (Fe1, Mg1) is not the same as the composition with the most kinetically ideal, fully reversible charge-storage behaviour (Fe5, Mg3): capacity and kinetic reversibility are correlated with dopant identity and loading but are not equivalent design targets, and both trends are quantified here within a single, consistent device platform.

Successive ionic layer adsorption and reaction (SILAR) is well suited to the fabrication of oxide films on conductive substrates. Alternating adsorption, rinsing and reaction steps limit precipitation in the bulk solution and allow the deposited amount to be controlled through cycle number. The method is inexpensive, vacuum free and compatible with complex substrate shapes.²⁰ Binder-free, SILAR-grown copper-oxide electrodes have recently been demonstrated in symmetric and asymmetric solid-state supercapacitor devices, supporting the general applicability of this deposition route to compact energy-storage architectures.²¹ Introducing a second cation into the precursor bath can modify nucleation, defect chemistry, coherent-domain growth, electronic transport and the electrolyte-accessible surface. The electrochemical outcome is therefore governed not simply by the amount of the added species, but by the coupled evolution of structure, morphology and interfacial kinetics.

Despite considerable interest in Cu₂O for electrochemical energy storage, the influence of controlled compositional modification on its structure, morphology, interfacial kinetics, and device performance remains insufficiently understood. Fe and Mg provide a useful comparison because they can influence the Cu₂O growth environment differently, with Fe affecting the electronic and redox response, while Mg can influence defect equilibria and particle growth. However, systematic correlations between these effects and the resulting electrochemical performance remain limited, particularly for directly deposited, binder-free solid-state devices. The present study therefore employs the same SILAR growth sequence and two-electrode configuration for both series, enabling a consistent comparison of how composition influences charge-transfer resistance, charge-storage kinetics, morphology, and areal capacitance without the ambiguity associated with different substrates or cell designs.

Here, pristine Cu₂O is compared with nominally Fe- and Mg-modified films prepared using 1, 3 and 5 at% precursor additions, denoted Fe1, Fe3, Fe5 and Mg1, Mg3, Mg5 according to the modifier and its nominal loading. X-ray diffraction (XRD), attenuated-total-reflectance Fourier-transform infrared spectroscopy (ATR-FTIR), field-emission scanning electron microscopy (FESEM) and energy-dispersive X-ray spectroscopy (EDX) are integrated with cyclic voltammetry (CV), galvanostatic charge–discharge (GCD), electrochemical impedance spectroscopy (EIS), scan-rate analysis and long-term cycling. The analysis addresses four questions: whether the cuprite-dominant phase is retained, how morphology evolves across the two series, whether charge-transfer resistance tracks areal capacitance, and what the scan-rate and cycling data support. The resulting full-device dataset is pronouncedly non-monotonic: Fe1 gives the largest areal capacitance, Mg1 the lowest fitted charge-transfer resistance, and higher Mg addition the strongest morphological reorganisation.

2. Experimental

2.1. Materials and sample preparation

Copper foil (0.1 mm thick, stated purity ≥99.9%) served as both the substrate and current collector. The foil was polished sequentially with 600-, 1200- and 2000-grit SiC paper and 0.3 µm alumina suspension. It was then etched in 1 M HCl for 300 s, sonicated in absolute ethanol and deionised water for 15 min each, and dried at 70 °C for 15 to 20 min.

The cationic bath contained 0.1 M CuCl₂·2H₂O and 0.05 M d-(+)-glucose in deionised water. A separate 0.2 M NaOH solution served as the anionic bath for the reaction step in the SILAR cycle. FeCl₃·6H₂O or MgCl₂·6H₂O was added to the cationic bath at nominal concentrations of 1, 3 or 5 at% relative to Cu. The resulting films are denoted Fe1, Fe3, Fe5, Mg1, Mg3 and Mg5; a bath without Fe or Mg produced the pristine Cu₂O reference.

Glucose acts as a mild reducing agent under alkaline conditions and favours conversion of adsorbed Cu(ii)-containing species to Cu(i) oxide.²⁰ An idealised overall reaction may be written as

2Cu²⁺ + C₆H₁₂O₆ + 5OH⁻ → Cu₂O + C₆H₁₁O₇⁻ + 3H₂O 1

The sequential adsorption and reaction steps confine nucleation to the substrate surface and suppress extensive precipitation in the bulk solution.

Deposition was conducted at 25 ± 2 °C for 20 SILAR cycles. Each cycle comprised 60 s in the cationic bath, a 30 s deionised-water rinse, 60 s in 0.2 M NaOH and a final 30 s rinse. The films were dried at 70 °C for 30 min and annealed at 350 °C for 90 min in static air. A schematic of the complete deposition sequence is provided in Scheme 1a.

Scheme 1. Fabrication of the binder-free Cu₂O-based symmetric solid-state supercapacitor. (a) Copper-substrate preparation, the four-step SILAR deposition cycle repeated for 20 cycles, and post-deposition oven drying and annealing; (b) preparation of the PVDF/NMP–LiCl/KOH gel electrolyte and face-to-face assembly of the composition-matched symmetric device.

2.2. Device assembly

Two composition-matched coated electrodes were assembled face-to-face with a poly(vinylidene fluoride) (PVDF)-based LiCl/KOH gel electrolyte. PVDF (1.0 g) was dissolved in N-methyl-2-pyrrolidone (NMP, 20 mL), after which LiCl (0.4 g) and KOH (0.1 g) were added and the mixture was stirred at 70 °C for 4 to 6 h. Complete dissolution of the polymer and homogeneous incorporation of the salts yield a bubble-free, optically homogeneous gel; incomplete dissolution instead gives a heterogeneous, partially gelled precursor and unreliable electrode–electrolyte contact. The gel was applied to both coated surfaces, the electrodes were assembled as a sandwich and the device was dried overnight at room temperature. Each electrode measures 2 × 2 cm, so the projected device area used for all areal calculations is 4 cm². The fabrication sequence is shown in Scheme 1b.

The electrolyte contains no added water at any stage: LiCl and KOH are dispersed directly in the PVDF/NMP solution, so the mobile species are Li⁺, K⁺, Cl⁻ and OH⁻ within an aprotic polymeric host rather than an aqueous or hydrogel medium. NMP is a strongly polar aprotic solvent for a wide range of inorganic salts, and the active films are annealed at 350 °C in static air before casting (Section 2.1), so no process water is carried into the assembled device. Because NMP boils at 202 °C and the devices were dried only at ambient temperature, residual NMP is retained and provides the medium for ion transport; the electrolyte is therefore an NMP-plasticised PVDF gel rather than a dry solid polymer electrolyte. Throughout this work the term binder-free describes the SILAR-grown Cu₂O active layer, which is deposited directly on the copper-foil current collector without polymeric binder or conductive-carbon additive; the gel electrolyte necessarily contains PVDF as the structural host for the LiCl/KOH ionic medium.

2.3. Characterization and electrochemical measurements

XRD patterns were recorded on a Bruker D8 Advance diffractometer with Cu Kα radiation (λ = 1.5406 Å) at 40 kV and 40 mA. Data were collected over 2θ = 10 to 90° using a 0.012° step and 1 s counting time per step. Coherent-domain sizes were estimated using the Scherrer relation, with K = 0.94 and instrumental broadening referenced to LaB₆.

D = Kλ/(β cos θ) 2

where D is the coherent-domain size, λ is the X-ray wavelength, β is the corrected full width at half maximum in radians and θ is the Bragg angle. ATR-FTIR spectra were recorded on a Bruker Vertex 70 using 64 scans, a resolution of 4 cm⁻¹ and a spectral range of 400 to 4000 cm⁻¹. Surface morphology and elemental composition were examined using a JEOL JSM-7600F FESEM equipped with EDX and operated at 10 kV.

Electrochemical measurements were performed with an Autolab PGSTAT204 in a two-electrode symmetric configuration. CV was recorded from −0.4 to +1.3 V at 20, 30, 50, 75 and 100 mV s⁻¹. GCD curves were obtained at 0.5, 1, 1.5 and 2 mA cm⁻². EIS was measured from 0.1 Hz to 100 kHz with a 10 mV perturbation. Spectra were fitted in ZView 3.5 by complex non-linear least squares with modulus weighting, repeated from randomised starting points to avoid local minima, to the circuit R_s–(R_ct∥CPE_dl)–W_s–(R_leak∥C_L), in which CPE_dl is a constant-phase element (CPE) representing the non-ideal double layer and W_s a finite-length Warburg element. Every constant-phase exponent was constrained to the physically meaningful range 0 < n ≤ 1, parameter standard errors were taken from the Jacobian at the solution, and the fit was assessed against the Bode magnitude and phase as well as the Nyquist plot. Long-term cycling stability was measured for pristine Cu₂O, Fe1, Fe3, Fe5, Mg1, Mg3 and Mg5 over 10 000 cycles at 4 mA cm⁻².

2.4. Electrochemical data analysis

All capacitance and energy values are reported for the complete symmetric device and normalised to the projected device area, the appropriate basis for a directly grown thin-film electrode of fixed footprint. CV and GCD were recorded over the identical −0.4 to +1.3 V (1.7 V) window for all seven devices, so that metrics derived from the two techniques remain directly comparable. The GCD current-density range of 0.5 to 2.0 mA cm⁻² was chosen to give fully resolved discharge transients with acceptably low ohmic (IR) drop relative to the measured potential window across all seven devices.

Device areal capacitance was calculated from the discharge branch:

C_A = IΔt/(AΔV) 3

where I is the device current, Δt is the discharge time, A is the projected device area and ΔV is the potential window. Eqn (3) is the time-averaged, device-level definition of areal capacitance used throughout. For the markedly non-linear discharge profiles obtained here, curve integration provides a complementary, potential-resolved measure of the stored energy, and both are reported. The areal energy and the time-averaged areal power were obtained from the discharge branch as

E_A = (J/3600)∫V(t)dt 4

P_A = (E_A × 3600)/Δt 5

where (J) is the current density (mA cm⁻²) and ∫V(t)dt is the integrated area under the discharge curve (V s), giving (E_A) in µWh cm⁻².

Capacitance retention and coulombic efficiency were calculated as

Retention (%) = (C_n/C₁) × 100 6

Coulombic efficiency (%) = (t_discharge/t_charge) × 100 7

where C₁ and C_n are the first-cycle and nth-cycle capacitances, and t_charge and t_discharge are the corresponding GCD times. Eqn (7) defines coulombic efficiency as the discharge-to-charge time ratio at constant current, evaluated over the identical potential window for both branches and with the IR drop treated consistently. The scan-rate dependence of the anodic peak current was evaluated with

i = av^b 8

Log i = b log v + log a 9

where i is the peak current, v is the scan rate and a and b are fitted constants; b = 1 corresponds to an idealised surface-controlled current and b = 0.5 to semi-infinite diffusion control.²² The potential-dependent current was further separated empirically using

i(V) = k₁v + k₂v^1/2 10

where k₁v describes the faster, scan-rate-proportional (capacitive) contribution and k₂v^1/2 describes the diffusion-associated contribution; the corresponding capacitive and diffusion contribution rates for all seven devices are analysed in Section 3.3.

3. Results and discussion

3.1. Structural and morphological characterization

The diffraction patterns of the pristine and modified films are dominated by reflections indexed to cubic cuprite Cu₂O (Powder Diffraction File, PDF 01-074-1230), as shown in Fig. 1a and b. No additional crystalline phase is resolved across either series. Retention of the cuprite-dominant pattern, within the sensitivity of laboratory powder XRD (typically a few volume percent for a crystalline secondary phase), shows that the SILAR and annealing conditions preserve the principal Cu₂O framework over the whole precursor-composition range.^14,20 Fe3 shows an enhanced relative intensity of the (110) reflection near 29.6°, which is attributed to Fe-induced modification of nucleation and crystal growth during deposition, promoting preferential growth of the (110) planes. Because no additional reflection appears at any loading and the (111), (200), (220) and (311) peaks remain indexed to cuprite Cu₂O throughout, this intermediate-loading intensification reflects texture rather than a change in phase purity.

Fig. 1. Structural evolution of pristine and modified Cu₂O films. (a) Fe-series XRD patterns with the expanded (111) region; (b) Mg-series XRD patterns with the expanded (111) region; and (c) the position of the (111) reflection as a function of nominal Fe or Mg addition, with the corresponding cubic lattice parameter on the right-hand axis. Error bars are the ±0.02° reproducibility obtained by reading pristine Cu₂O independently from both expanded panels.

The enlarged (111) region in Fig. 1a and b resolves the composition-dependent change in peak position and profile, which follows the evolving strain and defect environment of the lattice. Three converging observations identify Fe or Mg modification of the Cu₂O structure rather than the formation of a separate, diffraction-visible phase: the systematic (111) peak shift together with the composition-dependent coherent-domain size (Fig. 1c), the Cu–O bond-environment shift seen by ATR-FTIR (Fig. 2), and the co-located EDX Fe or Mg signal (Fig. 4), with no additional crystalline reflection at any loading up to 5 at%. The films are accordingly described as Fe- or Mg-modified Cu₂O throughout, following the same combination of techniques used for related Cu₂O-based supercapacitor electrodes.²³ XRD, ATR-FTIR and EDX are not directly sensitive to cation oxidation state or lattice site occupancy, so these data establish incorporation of the modifier into the film without discriminating between substitution in the cuprite lattice and a closely associated surface or interstitial defect population.

Fig. 2. ATR-FTIR spectra of pristine Cu₂O and the nominal Fe- and Mg-modified films. (a) Full-range spectra, with the shaded box marking the expanded region, and (b) the expanded low-wavenumber region, showing the composition-dependent Cu–O band shift relative to the pristine response near 606 cm⁻¹.

Fig. 4. EDX spectra of (a)–(d) pristine Cu₂O and the Fe-modified series and (e)–(h) pristine Cu₂O and the Mg-modified series, with the energy axis calibrated on the Cu Lα and Cu Kα lines. Dashed guides mark the expected positions of Fe Kα (6.404 keV) and Mg Kα (1.253 keV).

The (111) reflection shifts systematically with composition (Fig. 1c), moving from 36.74° for pristine Cu₂O to 36.69° for Fe1 and 36.45–36.47° for Fe3 and Fe5, and to 36.42–36.49° across the Mg series. Pristine Cu₂O appears in both expanded panels and is reproducible to ±0.02°, an order of magnitude below the shift itself. The corresponding cubic lattice parameter rises from 0.4234 nm to 0.4262–0.4269 nm, a small expansion in the direction expected when a larger cation is accommodated in the cuprite framework. Coherent-domain sizes obtained from the same reflections through the Scherrer relation (eqn (2)) are approximately 64 nm for pristine Cu₂O, increase at Fe1 and then refine to approximately 55 nm at Fe5, whereas the Mg series grows progressively to approximately 106 nm at Mg5. These are coherent crystallographic domain sizes rather than the particle diameters obtained by FESEM, and the two modifiers therefore alter nucleation and post-deposition crystallisation in different ways.

ATR-FTIR spectra retain the characteristic low-wavenumber Cu–O response and shift with composition relative to the pristine band near 606 cm⁻¹ (Fig. 2). Measured in the expanded region (Fig. 2b), the band minimum is displaced by approximately 12 cm⁻¹ to lower wavenumber for Fe5, is essentially unshifted for Fe3, and is displaced by 8 to 9 cm⁻¹ to higher wavenumber for Fe1, Mg1, Mg3 and Mg5. For a harmonic oscillator the vibrational wavenumber is related to the bond force constant k and the reduced mass µ by eqn (11)

ṽ = (1/2πc)(k/µ)^1/2 11

where c is the speed of light, so the measured shifts reflect a change in the effective Cu–O bonding environment produced by composition-dependent strain, defects and local coordination.¹⁶ Broad O–H-related stretching and bending bands are also observed. No additional band attributable to a crystalline secondary oxide is resolved, consistent with the cuprite-dominant XRD patterns (Section 3.1). Combined with EDX detection of Fe or Mg in the corresponding films (Fig. 4), the spectroscopic response establishes composition-dependent modification of the deposited Cu₂O films.

Representative FESEM images and the corresponding particle-size distributions, fitted with log-normal functions, are shown for all seven films in Fig. 3. Fe-modified films preserve a compact granular morphology with only minor variation in particle size, whereas increasing Mg content reorganises the surface, with Mg5 developing a more open and interconnected agglomerated texture. The pristine film consists of agglomerated Cu₂O particles of mean diameter 196 ± 12 nm (Fig. 3a, b and Table 2). Fe1 is slightly coarser at 205 ± 15 nm (Fig. 3c and d), while Fe3 and Fe5 give 182 ± 11 and 187 ± 13 nm (Fig. 3e–h). Particle sizes throughout this work are obtained from two-dimensional FESEM projections and log-normal fitting of the resulting histograms, and are therefore comparative across the present series rather than absolute. The continuity of the Fe series indicates that Fe addition changes nucleation and coherent-domain development without reconstructing the surface on a large scale.^7,13

Fig. 3. FESEM micrographs and corresponding particle-size distributions, fitted with log-normal functions, for (a) and (b) pristine Cu₂O, (c) and (d) Fe1, (e) and (f) Fe3, (g) and (h) Fe5, (i) and (j) Mg1, (k) and (l) Mg3 and (m) and (n) Mg5; a 500 nm scale bar is shown in each micrograph.

Table 2. Central structural and electrochemical metrics. Areal capacitance is the device value at the lowest current density, 0.5 mA cm⁻²; areal energy and average areal power are calculated over ΔV = 1.7 V from eqn (4) and (5). Charge-transfer resistances are the fitted values of Table 4.

Sample	Particle diameter (nm)	R ct (Ω)	C A (mF cm−2)	E A (µWh cm−2)	P A (µW cm−2)
Pristine Cu2O	196 ± 12	(3.9 ± 0.7) × 104	12	1.6	146
Fe1	205 ± 15	759 ± 13	129	9.7	80
Fe3	182 ± 11	231 ± 4	14	0.4	35
Fe5	187 ± 13	20.0 ± 0.4	8	0.5	76
Mg1	162 ± 10	18.5 ± 0.4	80	0.8	11
Mg3	190 ± 14	256 ± 11	4	0.3	80
Mg5	135 ± 9	103 ± 21	14	0.9	70

Mg modification produces a stronger morphological evolution (Fig. 3i–n). Mg1 contains smaller particles (162 ± 10 nm, Fig. 3i and j), Mg3 returns to 190 ± 14 nm (Fig. 3k and l), and Mg5 gives the smallest fitted diameter (135 ± 9 nm) together with the most open, interconnected agglomerated texture in the series (Fig. 3m and n). The open Mg5 texture is expected to favour electrolyte access, although the electrochemical response also depends on interfacial resistance and on the population of accessible redox sites (Section 3.3). Larger coherent domains in XRD combined with smaller apparent surface particles at Mg5 point to fused or differently oriented domains within a hierarchically agglomerated film, so the morphology reflects both crystallographic growth and the assembly of particles at the substrate. The stronger response to Mg than to Fe follows from the different structural roles of the two cations during growth: Mg²⁺, with a harder-acid character and a coordination preference distinct from Cu⁺, segregates more readily toward grain boundaries and surfaces and acts as a growth-modifying species that perturbs nucleation and coarsening, whereas Fe³⁺ is more readily accommodated within the cuprite framework with less disruption to particle coarsening.

EDX spectra confirm Cu and O in every film and show Fe or Mg signals exclusively in the corresponding modified series, with no cross-contamination between them (Fig. 4). The measured contents rise with nominal precursor addition: 1.7, 4.5 and 5.9 at% Fe for Fe1, Fe3 and Fe5, and 1.3, 3.6 and 4.7 at% Mg for Mg1, Mg3 and Mg5 (Table 1). The measured Cu : O ratio falls correspondingly, from 2.16 for pristine Cu₂O to 1.80 for Fe5 and 1.87 for Mg5. These ratios are comparative within the present series rather than absolute stoichiometries, because light-element quantification by EDX in a thin film on a copper substrate depends on the interaction volume, the substrate contribution, the beam energy and the measurement geometry. The spectra of Fig. 4 provide spatially resolved confirmation of modifier incorporation that is independent of, and consistent with, the composition trend of Table 1.

Table 1. Semi-quantitative elemental composition of the films from EDX. Cu and O contents are normalised to 100% together with Fe or Mg; an em dash indicates a content below the EDX detection limit. The ratios are comparative within this series rather than absolute stoichiometries.

Sample	Cu (at%)	O (at%)	Fe (at%)	Mg (at%)	Cu : O ratio
Pristine Cu2O	68.4	31.6	—	—	2.16
Fe1	65.2	33.1	1.7	—	1.97
Fe3	63.1	32.4	4.5	—	1.95
Fe5	60.5	33.6	5.9	—	1.80
Mg1	66.2	32.5	—	1.3	2.04
Mg3	64.5	31.9	—	3.6	2.02
Mg5	62.1	33.2	—	4.7	1.87

3.2. Electrochemical performance and charge-storage kinetics

The symmetric devices were evaluated over a 1.7 V window from −0.4 to +1.3 V. Their CV profiles are distorted quasi-rectangular loops carrying broad redox features rather than ideal rectangles (Fig. 5a–g), the response characteristic of mixed capacitive and faradaic charge storage coupled to distributed transport resistance.^4–6,24 Current rises continuously with scan rate while the broad features remain visible from 20 to 100 mV s⁻¹. Fe1 encloses the largest voltammetric area of the Fe series; its broad redox envelope and increased current indicate a larger population of electrochemically accessible sites together with faster interfacial transport. The Mg series behaves differently: Mg1 gives the largest response, Mg3 is strongly suppressed and Mg5 partially recovers. The recovery at Mg5 parallels the open morphology seen by FESEM, while the larger response of Mg1 shows that surface architecture alone does not set the stored charge. Within the Fe series, Fe5 reaches a higher absolute peak current than Fe1 at the fastest scan rate, consistent with its larger power-law exponent (b = 0.51, Section 3.3), yet Fe1 encloses the larger total voltammetric area, and therefore the greater integrated charge, at every scan rate. Peak current and enclosed area are related but distinct descriptors, and it is the latter that determines the higher device areal capacitance of Fe1 (Table 2).

Fig. 5. Electrochemical response of the symmetric solid-state devices. (a)–(g) CV curves of the pristine Cu₂O, Fe1, Fe3, Fe5, Mg1, Mg3 and Mg5 devices, recorded from 20 to 100 mV s⁻¹.

The choice of a 1.7 V window follows from the composition of the electrolyte. Because it contains no added water, the ionic medium being Li⁺, K⁺, Cl⁻ and OH⁻ immobilised in a PVDF/NMP matrix rather than a bulk aqueous or hydrogel phase (Section 2.2), the 1.23 V thermodynamic decomposition voltage of water is not the relevant ceiling for this cell. The accessible window is bounded instead by the intrinsic anodic and cathodic stability of NMP, possible oxidation of Cl⁻ at the positive electrode, the electrochemical and thermal stability of the PVDF host, and reduction of Li⁺ and K⁺ at the negative electrode. The window was established empirically as the range over which all seven devices cycle reversibly and reproducibly, and two features of the data support it. At the anodic vertex (+1.3 V) the current remains bounded and comparable in magnitude to the redox currents elsewhere in the voltammogram, and the loops close rather than developing the steep, near-vertical rise of a solvent-decomposition wave. Over 10 000 cycles the coulombic efficiency of every device settles onto a stable plateau rather than declining continuously (Section 3.4), which is what a progressive electrolyte-consuming reaction at the vertex would produce. Direct determination of the stability limit by linear-sweep voltammetry to the point of catastrophic current rise, together with chronoamperometric polarisation decay, would place a quantitative bound on the window and is a natural extension of this work.

The GCD curves show composition-dependent discharge times and varying degrees of curvature (Fig. 6a–g). At 0.5 mA cm⁻², pristine Cu₂O delivers 12 mF cm⁻². Fe1 reaches 129 mF cm⁻², a 10.7-fold increase, and Mg1 reaches 80 mF cm⁻², 6.7 times the pristine value. Further addition is not beneficial: Fe3 and Fe5 give 14 and 8 mF cm⁻², and Mg3 and Mg5 give 4 and 14 mF cm⁻². The CV and GCD data therefore show the same non-monotonic composition dependence.

Fig. 6. Electrochemical response of the symmetric solid-state devices. (a)–(g) GCD curves of the same devices in the same order, measured from 0.5 to 2 mA cm⁻².

Over the same 1.7 V interval the areal energy densities are 1.7, 9.7, 0.5, 0.5, 0.8, 0.3 and 0.9 µWh cm⁻² for pristine Cu₂O, Fe1, Fe3, Fe5, Mg1, Mg3 and Mg5 (Table 2), so Fe1 stores the most energy of the series. The enhancement at low addition is consistent with modifier-induced lattice modification and defect chemistry in Cu₂O, which alter the electronic structure, facilitate charge transfer and improve access to electrochemically active sites.^4,6,24 At higher loading, increased lattice distortion and defect density hinder charge transport and reduce the availability of active sites, which accounts for the weaker response of Fe3, Fe5 and Mg3. The structural and kinetic data trace this directly: the capacitance and energy maxima at low addition coincide with a low fitted charge-transfer resistance only for Mg1 (18.5 Ω, Tables 2 and 4), whereas Fe1 reaches its maximum despite a comparatively large charge-transfer resistance (759 Ω), so accessible charge and interfacial kinetics are not set by the same structural factor. Neither maximum is accompanied by a large increase in coherent-domain size, consistent with a gain in carrier density and interfacial kinetics at low loading that is not yet offset by lattice strain; at higher loading the larger coherent-domain and R_ct changes (Table 4) point to distortion and scattering that outweigh any further electronic benefit. The series was limited to nominal 1, 3 and 5 at% addition in order to bracket low, intermediate and high loading, and because 1 at% is already optimal for both modifiers among the levels tested, the true optimum may lie below 1 at%; finer composition mapping between 0.1 and 1 at% is the natural next step. The partial recovery at Mg5 shows that the electrochemical behaviour is governed by the balance between modifier concentration, defect structure and charge-transport kinetics. Literature values for Cu-based materials must be compared with care, because differences in electrode configuration, normalisation basis, electrolyte and voltage window produce substantial variation in reported energy density (Table 3).^7,13,14,16

Table 4. Equivalent-circuit parameters from complex non-linear least-squares fitting of the raw impedance spectra to the circuit of Fig. 7e, R_s–(R_ct∥CPE_dl)–W_s–(R_leak∥C_L). R_s, R_ct, R_W and R_leak are in Ω, CPE_dlQ in S sⁿ, T_W in s, C_L in F, and n is dimensionless. σ_W = R_W/(2T_W)^1/2 is the derived Warburg coefficient in Ω s^−1/2. Uncertainties are one standard error from the fit. Fit quality is given as the reduced chi-squared together with the root-mean-square deviation of the fitted impedance magnitude and phase angle, so the circuit is judged in the Bode representation as well as the Nyquist one.

Parameter	Cu2O	Fe1	Fe3	Fe5	Mg1	Mg3	Mg5
R s (Ω)	1299.12 ± 352.06	34.56 ± 0.41	22.22 ± 0.11	3.33 ± 0.02	2.54 ± 0.02	8.11 ± 0.03	69.24 ± 17.62
R ct (Ω)	39 058.7 ± 7199.9	759.3 ± 12.6	231.1 ± 3.7	20.0 ± 0.4	18.5 ± 0.4	256.2 ± 11.1	103.2 ± 21.4
CPEdlQ (S sn)	1.21 × 10−5	1.09 × 10−4	4.31 × 10−5	2.24 × 10−4	1.42 × 10−4	3.69 × 10−4	1.88 × 10−5
CPEdln	0.300 ± 0.034	0.650 ± 0.002	0.767 ± 0.005	0.845 ± 0.008	0.895 ± 0.008	0.730 ± 0.006	0.481 ± 0.077
R W (Ω)	4591.2 ± 336.3	10.3 ± 0.4	279.2 ± 7.7	24.4 ± 0.7	22.7 ± 0.9	33.7 ± 9.3	1435.0 ± 56.2
T W (s)	9.63 × 10−6	4.49 × 10−6	2.35 × 10−1	1.66 × 10−1	9.73 × 10−2	5.74 × 10−1	6.34 × 10−2
σ W (Ω s−1/2)	1.05 × 106	3.43 × 103	407	42.3	51.5	31.5	4.03 × 103
R leak (Ω)	7464.3 ± 2114.3	302.6 ± 10.3	463.7 ± 11.9	35.0 ± 1.4	22.7 ± 0.8	84.9 ± 6.3	1819.6 ± 52.4
C L (F)	1.48 × 10−4	1.38 × 10−4	2.50 × 10−3	3.24 × 10−2	1.83 × 10−2	2.02 × 10−2	8.07 × 10−5
χ 2 (reduced)	1.4 × 10−3	3.7 × 10−5	2.5 × 10−4	5.0 × 10−4	5.0 × 10−4	1.4 × 10−4	4.1 × 10−4
|Z| rms dev. (%)	4.59	0.55	1.39	1.91	1.49	1.12	1.77
Phase rms dev. (°)	1.21	0.35	0.95	1.37	1.51	0.66	1.22

Table 3. Comparison of selected Cu-based supercapacitor devices, grouped by measurement configuration. The upper group lists two-electrode, full-device results, the appropriate basis for comparison with this work; the lower group lists three-electrode half-cell values, included for qualitative context only and not comparable with full-device metrics.

Material	Configuration	Electrolyte	Metric	Ref.
Two-electrode, full-device values–the appropriate basis for comparison with this work
PPy–Cu2O–MoO3/GO/GFE	Symmetric solid-state	PVA/H2SO4 gel	596.5 mF cm−2 at 1 mA cm−2	25
Cu2O/Cu/cellulose	Symmetric two-electrode device	1 M KOH (aqueous)	213 mF cm−2 at 1.9 A g−1	26
CuO/CuO	Symmetric two-electrode device	1 M KOH (aqueous)	983.33 F g−1 at 20 mV s−1	27
Fe1-modified Cu2O (this work)	Two-electrode symmetric device	PVDF/NMP–LiCl/KOH gel	129 mF cm−2 at 0.5 mA cm−2	—
Mg1-modified Cu2O (this work)	Two-electrode symmetric device	PVDF/NMP–LiCl/KOH gel	80 mF cm−2 at 0.5 mA cm−2	—
Three-electrode (half-cell) literature values–shown for qualitative context only; not directly comparable with the full-device values above
CuO/Cu2O/Cu on Ni foam	Three-electrode half-cell	1 M KOH (aqueous)	878 F g−1 at 5 mA cm−2	13
Cu2O microspheres	Three-electrode half-cell	2 M KOH (aqueous)	144 F g−1 at 0.1 A g−1	14
O–Cu2O/rGO composite	Three-electrode half-cell	Alkaline aqueous	525 F g−1 at 1 A g−1	16
CuCo2O4@NiCo2O4	Three-electrode half-cell	Aqueous alkaline	2.59 F cm−2 at 10 mA cm−2	7

3.3. Capacitive and diffusion-controlled charge-storage contributions

The relative contributions of surface-controlled and diffusion-controlled charge storage were resolved by two complementary treatments, both applied to the discharge-branch areal capacitance at each scan rate. The first is the Trasatti extrapolation, which uses the two standard linearisations of the same dataset. Plotting 1/C against v^1/2 (Fig. 7a) and extrapolating to v^1/2 → 0, the infinitely slow scan-rate limit, gives the total capacitance C_total, at which both the readily accessible outer surface and the less accessible inner surface contribute. Plotting C against v^−1/2 (Fig. 7b) and extrapolating to v^−1/2 → 0, the infinitely fast limit, gives the outer, readily accessible surface capacitance C_outer, since only the most accessible surface can respond at very high scan rate; the difference C_total − C_outer is the inner, less accessible contribution. The second treatment is the current deconvolution of eqn (10), which partitions the measured current at each potential into a term proportional to v (surface-controlled) and a term proportional to v^1/2 (diffusion-controlled); the contribution rates of Fig. 6c come from this deconvolution integrated over the CV window and are a separate quantity from the Trasatti intercepts. Both Trasatti linearisations are linear for all seven devices across 20 to 100 mV s⁻¹, consistent with the mixed response already indicated by the quasi-rectangular CV loops with superimposed broad redox features (Section 3.2). Eqn (10) is an empirical kinetic partition rather than a thermodynamic assignment of mechanism: a current scaling linearly with v is termed surface-controlled, which is not in itself evidence that the corresponding charge is stored by double-layer formation rather than by a fast, surface-confined faradaic process. The two treatments are reported together because they probe different aspects of the same response.

Fig. 7. Capacitive and diffusion-controlled charge-storage analysis of the symmetric devices. (a) 1/C versus v^1/2 and (b) C versus v^−1/2, the two Trasatti linearisations, for pristine Cu₂O and the Fe- and Mg-modified series; and (c) the capacitive and diffusion contribution rates for all seven devices, obtained from the current deconvolution of eqn (10).

The corresponding capacitive and diffusion contribution rates, obtained from the k₁v and k₂v^1/2 terms of eqn (10) integrated over the CV window, are summarised in Fig. 7c. Pristine Cu₂O contributes 74% of its charge through the surface-controlled term and 26% through the diffusion-controlled term. Fe modification increases the capacitive fraction sharply at low addition, to 94% for Fe1, before declining to 71% for Fe3 and recovering to 88% for Fe5. The Mg series shows the opposite low-concentration trend: the capacitive fraction falls to 63% for Mg1, recovers to 93% for Mg3, and falls again to 59% for Mg5, the most diffusion-limited device in the study (41% diffusion contribution).

The high capacitive fraction obtained for Fe1 is not explained by its interfacial kinetics: its fitted charge-transfer resistance is one of the larger values of the series (759 Ω, Section 3.4), while it retains the largest areal capacitance (129 mF cm⁻², Section 3.2): a large surface-controlled charge contribution and fast interfacial charge transfer are therefore separable properties in this series rather than mutually reinforcing ones. Conversely, the elevated diffusion contribution at Mg5 (41%) is consistent with its open, interconnected agglomerated morphology (Section 3.1), in which longer and more tortuous ion-transport pathways through the film favour a more diffusion-limited response even though the same morphology may also increase the electrolyte-accessible surface area, which FESEM alone cannot establish.

The contribution-rate ranking does not follow the anodic peak-current power-law exponent b obtained from eqn (8), which is 0.24, 0.39, 0.26, 0.51, 0.30, 0.23 and 0.35 for pristine Cu₂O, Fe1, Fe3, Fe5, Mg1, Mg3 and Mg5. The two quantities are not expected to coincide: b describes the scan-rate scaling of the anodic peak current alone, whereas the contribution rate decomposes the entire integrated voltammetric current at a fixed scan rate and is not bounded by the idealised 0.5 to 1.0 range. Exponents below the semi-infinite-diffusion limit of 0.5, as obtained here throughout and most markedly for Mg3 (b = 0.23), are reported for systems with finite-size or thin-film diffusion geometry, overlapping redox processes, or distributed relaxation of the kind already established for these devices by the constant-phase-element behaviour in EIS (Section 3.4); they are diagnostic of mixed, non-ideal peak kinetics.^28,29 The clearest illustration is Fe5, which combines the largest b of the series (0.51) with a high capacitive contribution rate (88%), so its peak current and its total charge are governed by different rate-limiting steps.

3.4. Electrochemical impedance and cycling stability

Electrochemical impedance spectroscopy separates the resistive, diffusive and capacitive contributions of the assembled devices (Fig. 8). Every spectrum shows the same three features: a depressed high-frequency arc, a sloping mid-frequency region, and a low-frequency arc that closes back onto the real axis rather than rising vertically. The circuit of Fig. 8e contains one element for each. R_s is the series resistance of the electrolyte, current collectors and contacts. R_ct in parallel with the constant-phase element CPE_dl describes charge transfer across a non-ideal electrode/electrolyte interface. W_s is a finite-length Warburg element, W_s = R_W tanh(x)/x with x = (jωT_W)^1/2, describing ion diffusion over a finite path length in the film. The low-frequency branch places the device capacitance C_L in parallel with a leakage resistance R_leak. Both of these elements are demanded by the data: omitting the Warburg element leaves a systematic residual in exactly the mid-frequency decade where the 45° response occurs and raises the reduced chi-squared by factors of 1.1 to 14 across the seven devices, while a purely capacitive termination cannot reproduce a low-frequency arc that closes onto the real axis. That closure is itself informative: it shows that every cell carries a finite direct-current (DC) conduction path in parallel with its capacitance, which a leakage-current and self-discharge measurement would quantify directly.

Fig. 8. Electrochemical impedance response of the symmetric solid-state devices. (a) Nyquist plots of the Fe-modified series and (b) of the Mg-modified series, each with an inset enlarging the low-impedance region; (c) pristine Cu₂O on its own scale, its impedance being roughly thirty times larger; (d) impedance magnitude of all six modified devices across the full frequency range, showing the quality of the fit decade by decade; and (e) the equivalent circuit, R_s–(R_ct∥CPE_dl)–W_s–(R_leak∥C_L). Symbols are measured data and solid lines the fitted spectra.

Fitting was performed on the raw spectra by complex non-linear least squares with modulus weighting and multi-start optimisation, with the frequency ordering of the three processes imposed as a constraint so that the arcs cannot exchange roles. The fit reproduces the impedance magnitude to within 0.55–4.59% and the phase angle to within 0.35–1.51° over the full 0.1 Hz to 100 kHz window (Table 4 and Fig. 8, 9), and every parameter of every device, pristine Cu₂O included, is determined to better than about 21% of its value. The series resistance falls from 1299 Ω for pristine Cu₂O to 2.5–69 Ω for the modified devices. The charge-transfer resistance falls by two to three orders of magnitude, from (3.9 ± 0.7) × 10⁴ Ω for pristine Cu₂O to 759 ± 13 Ω for Fe1, 231 ± 4 Ω for Fe3 and 20.0 ± 0.4 Ω for Fe5, and to 18.5 ± 0.4 Ω for Mg1, 256 ± 11 Ω for Mg3 and 103 ± 21 Ω for Mg5. The constant-phase exponent rises from 0.30 for pristine Cu₂O to 0.48–0.90 for the modified films, so the interfaces also become more nearly ideally capacitive.

Fig. 9. Bode plots of the symmetric devices. (a) Phase angle and (b) impedance magnitude for pristine Cu₂O and the Fe-modified series; (c) phase angle and (d) impedance magnitude for pristine Cu₂O and the Mg-modified series. Symbols are measured data and solid lines the same fit as in Fig. 7, so the circuit is validated against the phase angle and the impedance magnitude as well as against the Nyquist representation.

The Warburg element adds a description of ion transport. Taking σ = R_W/(2T_W)^1/2 as the Warburg coefficient, diffusion limitation falls by four orders of magnitude on modification, from 1.0 × 10⁶ Ω s^−1/2 for pristine Cu₂O to 31 Ω s^−1/2 for Mg3, with Fe5 (42), Mg1 (52) and Fe3 (4.1 × 10²) also far below the unmodified film, while Fe1 (3.4 × 10³) and Mg5 (4.0 × 10³) remain the most diffusion-limited of the modified devices. Modification therefore improves interfacial kinetics and ion transport together, but not in the same proportion, and neither improvement tracks the stored charge. Mg1 and Fe5 combine the lowest charge-transfer resistances of the series with among the lowest Warburg coefficients, yet Fe1, which stores the most charge, carries both a high R_ct and a high σ. The impedance response and the capacitance are governed by different structural factors, and no single fitted impedance parameter predicts the capacitance ranking; that separation, rather than any one resistance value, is the substantive result of the impedance analysis.

The Bode representation supports the same picture (Fig. 9). The phase-angle spectra show broad relaxation features for the modified samples, with maxima of approximately 42° for Fe1, 42° for Fe3 and 41° for Fe5, and 45°, 48° and 42° for Mg1, Mg3 and Mg5. Pristine Cu₂O, by contrast, remains near 12 to 15° across almost the entire measured range and rises only above 10 kHz, so it presents no resolvable relaxation peak within the window. These broad phase responses indicate non-ideal capacitive behaviour and distributed relaxation, consistent with the constant-phase element in the equivalent circuit. The |Z| spectra show a progressive decrease in impedance magnitude for the modified samples over most of the measured range. Across all seven devices the high-frequency intercept (Fig. 8b and d) reflects the series and contact resistance, the mid-frequency phase maximum reflects the dominant capacitive relaxation time constant, and the low-frequency phase angle and magnitude trace the transition toward resistive and diffusive behaviour; the near-flat, low phase angle of pristine Cu₂O is consistent with its poorly defined capacitive response.

The fitted series resistance is likewise substantially higher for pristine Cu₂O than for any modified sample (Table 4), which is attributed to poorer electrolyte wetting and higher contact resistance at the more compact, less electrolyte-accessible unmodified film identified by FESEM (Section 3.1) – the same improvement in interfacial accessibility that lowers R_ct on Fe or Mg modification. The comparison between Fe1 and Mg1 is the most informative in the dataset. Their fitted charge-transfer resistances differ by more than an order of magnitude, 759 Ω for Fe1 against 18.5 Ω for Mg1, yet it is Fe1 that stores the larger charge. Charge-transfer resistance alone therefore does not determine charge-storage capability, and the higher capacitance of Fe1 must arise from a larger population of electrochemically accessible redox-active sites. Compositional modification improves interfacial charge transfer, but the overall charge-storage performance depends on several factors acting together.

Long-term cycling traces for all seven devices—pristine Cu₂O, Fe1, Fe3, Fe5, Mg1, Mg3 and Mg5—are presented in Fig. 10. All seven devices retain approximately 95–99% of their capacitance after 10 000 cycles. Fe1, Fe5, and Mg3 show the highest retention of approximately 99%, while pristine Cu₂O, Mg1, and Mg5 retain about 98% and Fe3 retains about 95%. The high capacitance retention indicates good structural persistence of the Cu₂O-based devices during prolonged cycling. Compositional modification by Fe or Mg is consistent with changes in the local Cu–O coordination environment (Section 3.1), which may contribute to the improved structural and interfacial charge-transport behaviour observed here.²⁰

Fig. 10. Long-term galvanostatic cycling stability of the symmetric devices at 4 mA cm⁻² over 10 000 cycles: capacitance retention (left axis) and coulombic efficiency (right axis) for the pristine Cu₂O, Fe1, Fe3, Fe5, Mg1, Mg3 and Mg5 devices (panels (a)–(g), respectively).

Coulombic efficiency depends on composition far more strongly than capacitance retention does. Fe5 and Mg3 approach 97%, whereas pristine Cu₂O, Fe3 and Fe1 end near 76%, 76% and 75%, and Mg1 and Mg5 reach only 61% and 67%. The high efficiency of Fe5 and Mg3 reflects more reversible charge-storage reactions and reduced irreversible charge loss, consistent with their more favourable charge-transfer characteristics. Higher or non-optimal loading correlates with lower coulombic efficiency across the series, indicating a greater extent of irreversible charge loss per cycle; identifying the specific loss pathways would require post-cycling compositional and spectroscopic analysis.

The form of the efficiency curves is as informative as their terminal values. For every device, including Mg1 and Mg5, coulombic efficiency declines over the first several hundred cycles and then holds an essentially constant plateau through the remaining cycles, while capacitance retention over the same window stays high and stable at 95 to 99%. A cumulative degradation process would instead produce a progressively declining retention, which is not observed. The data are therefore consistent with a stable but incomplete charge-recovery process that reaches a constant, composition-dependent extent on every cycle, giving high capacitance retention alongside composition-dependent irreversible charge loss. Fe1 and Mg1 accordingly offer the largest accessible charge in this series, while Fe5 and Mg3 offer the more reversible charge storage at lower absolute capacitance; Fe5 provides the best balance within the Fe series and Mg3 within the Mg series.

3.5. Composition-dependent structure–property correlations

Fig. 11 compares the central electrochemical descriptors across the full composition series. Areal capacitance, charge-transfer resistance and Warburg coefficient all change sharply at 1 at% addition and then behave non-monotonically. Fe1 is the capacitance optimum, Mg1 the resistance optimum, Mg3 the diffusion optimum and Fe5 the reversibility optimum, so no single structural or kinetic parameter reproduces the complete performance ranking. Four distinct descriptors are used in this work and they need not order the compositions identically: the absolute areal capacitance, the total accessible charge per unit area, is highest for Fe1 (129 mF cm⁻², Table 2); the capacitive fraction of the total charge at fixed scan rate is also highest within the Fe series for Fe1 (94% against 88% for Fe5, Fig. 6c); the anodic peak-current power-law exponent b is largest, and closest to capacitive scaling, for Fe5 (0.51, Section 3.3); and long-term coulombic efficiency is highest for Fe5 (97% against 75% for Fe1, Section 3.4). Fe1 therefore stores the largest total charge, while Fe5 stores less charge with more kinetically ideal and more fully reversible characteristics. These are complementary observations once the specific meaning of each metric is kept distinct.

Fig. 11. Composition-dependent device metrics. (a) Areal capacitance at 0.5 mA cm⁻²; (b) fitted charge-transfer resistance and (c) Warburg coefficient, both on logarithmic scales; and (d) terminal coulombic efficiency after 10 000 cycles, for pristine Cu₂O and the Fe- and Mg-modified series. The four panels do not rank the compositions in the same order, which is the central observation of this work.

The time-averaged areal power during discharge follows from the same dataset through eqn (5). For an ideally linear, triangular discharge it would reduce to P_A = IΔV/2 and would be identical for all seven devices at a given current, but the discharge profiles here are markedly non-linear, so the measured values differ between compositions: at 0.5 mA cm⁻² they span 11 to 146 µWh-equivalent per µW per cm² (Table 2), with the highest value for pristine Cu₂O, whose short discharge concentrates a small stored energy into a short time. Average power at a single current density is thus dominated by the ratio of integrated discharge energy to discharge time and is a weak discriminator between devices. A genuine device-to-device power comparison, and the corresponding Ragone representation, requires a wider current-density sweep than the present dataset covers.^30,31

The integrated data reveal two distinct composition-dependent pathways. Low Fe addition produces the largest increase in electrochemically accessible charge together with a substantial reduction in interfacial resistance, while the comparatively stable granular morphology indicates that local electronic and redox changes dominate. Mg modification acts more strongly on film growth and surface architecture, yet its best capacitance occurs at Mg1 rather than at the most open Mg5 morphology, showing that accessible surface, electronic transport and active-site density must be balanced simultaneously.^1,4,11

4. Conclusions

Binder-free Cu₂O films with nominal Fe or Mg additions of 1, 3 and 5 at% were grown directly on copper foil by a 20-cycle SILAR process and evaluated as complete symmetric solid-state devices. The cuprite-dominant phase is retained across both series, while the (111) reflection, the coherent-domain size, the Cu–O bonding environment and the surface morphology evolve systematically with composition: Fe-modified films stay compact and granular, whereas Mg modification drives a stronger reorganisation of particle size and surface architecture. The best response occurs at the lowest addition. At 0.5 mA cm⁻², Fe1 delivers 129 mF cm⁻² and 9.7 µWh cm⁻² and Mg1 delivers 80 mF cm⁻² and 0.8 µWh cm⁻², against 12 mF cm⁻² and 1.7 µWh cm⁻² for pristine Cu₂O. The fitted charge-transfer resistance is 759 Ω for Fe1 and 18.5 Ω for Mg1, against (3.9 ± 0.7) × 10⁴ Ω for pristine Cu₂O, and the lowest values of the series belong to Mg1 and Fe5 rather than to the two highest-capacitance devices. The Warburg coefficient falls by four orders of magnitude on modification, from 1.0 × 10⁶ Ω s^−1/2 for pristine Cu₂O to 31 Ω s^−1/2 for Mg3. All seven devices retain 95 to 99% of their capacitance over 10 000 cycles, while coulombic efficiency remains strongly composition-dependent, 97% for Fe5 and Mg3 against 75% for Fe1 and 61% for Mg1.

The central result is that these descriptors do not rank the compositions identically. Mg1 has the lower charge-transfer resistance yet Fe1 stores the larger charge; Mg3 is the least diffusion-limited yet is among the lowest in capacitance; and the most reversible compositions are not the highest-capacitance ones. Interfacial kinetics, ion transport, accessible charge and reversibility therefore behave as correlated but non-equivalent design targets rather than as a single figure of merit, and Fe and Mg access distinct, complementary regions of that landscape. Direct SILAR growth of Fe- and Mg-modified Cu₂O is accordingly established here as a compositionally tunable, binder-free platform for compact solid-state energy storage, in which the modifier and its loading can be chosen against the descriptor that matters for a given application.

Author contributions

Zikra Javed: investigation, methodology, electrochemical characterisation and writing – original draft. Maria Bibi: structural characterisation and investigation. Yousra Noor: data curation and formal analysis. Baseena Sardar: investigation. Majid Khan: conceptualisation, supervision, project administration, funding acquisition and writing – review and editing. Inam Ullah, Haseeb Jamal, Farooq Khan and Tarfah Al-Warhi: contributed to discussion, validation and review of the manuscript. All authors reviewed and approved the final manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. All data supporting the findings of this study are provided within the main text.