Comparative study on surface oxygenation and widening of carbon fibers made electrochemical electrodes

Electrochemical capacitors produce a greater energy and power density and lie between electrostatic capacitors and batteries in the Ragone plot.¹ The report indicates two mechanisms involved in energy storage at solid–liquid interfaces where the faradaic redox reaction (FROR) yields a larger specific capacitance (Cs) than does electric double layer charging (EDLC). Accordingly, a recent study focuses very much on metal compounds with high oxidation states and electrical conductivity.^2–6 The carbon based electrodes are governed by EDLC and, due to their environmental friendliness, high retention, chemical stability, and low production cost, maintain prevailing in current markets. Two methods are widely adopted to improve the performance of carbon electrodes, including asymmetric configurations and the creation of micro-pores; the former is expressed as 1/Cs_(O) = 1/Cs_(P) + 1/Cs_(N), where Cs_(O) is overall Cs and, the Cs_(C) and Cs_(A) denote Cs contributions by cathode and anode made from different materials. Accordingly, the Cs_(C) ≠ Cs_(A) and Cs_(O) are determined by the one with lower Cs.⁷ Electrochemical cells with symmetric configuration, on the other hand, give Cs_(O) = 1/2Cs_(C) + 1/2Cs_(A) where the cathodes and anodes are made of same materials; thus, Cs_(C) = Cs_(A). For lithium-ion batteries, the half-reaction at the graphite anode is LiC₆ ↔ Li⁺ + C₆ + e⁻ so electrodes are often roughened to increase the number of C₆ units (i.e., widening of surface area, S_A). The creation of porous structures in capacitor electrodes, however, does not necessarily lead to Cs improvement because (i) cavities cannot induce charges and (ii) charge density at flat regions is severely limited by Coulomb repulsion [Figs. 1(a) and 1(b)].⁷ Alternatively, Cs may be promoted through polarization according to the following equation: Cs = εS_A/d, where d and ε denote dielectric thickness and permittivity [Figs. 1(b) and 1(c)].

In this work, both Cs ∝ ε and Cs ∝ S_A relations are studied and compared to understand which one dominates over Cs improvement. The carbon fiber cloths (CFCs) are used as electrode matrices for the following: (i) flexible, (ii) processing with ease, and (iii) resistance to oxidation and corrosion.^8–12 The Cs is then improved by coating carbon nanotubes (CNTs) and plasma treatments; the former is a one-dimension conductor made of rounded graphite sheets. Accordingly, CNT coating extends hexagonal lattices and S_A is widened (i.e., Cs ∝ S_A).¹³ Plasma treatments, on the other hand, create oxygenated lattices where charges accumulate, thus forming polarized states near to Fermi level (i.e., Cs ∝ ε).¹⁴ The results indicate ε > S_A in contribution to Cs, and the underlying mechanism is discussed.

Various CFCs based electrodes are prepared, including as-purchased (A_p-CFCs) [supplementary material, Figs. S1(a)–S1(c)], heat treated (H_T-CFCs), CNTs-coated (CNTs-CFCs), and plasma treated (P_T-CFCs), where CFCs are made from polyacrylonitrile (PAN) type and are purchased from suppliers (95% purity, Homytech and Conjutek, Taiwan). First, H_T-CFCs are made by heating A_P-CFCs to 1000 °C in an N₂ atmosphere for 1 h [supplementary material, Fig. S1(d)]. Second, CNTs-CFCs are produced in three steps: (i) multi-walled CNTs (MWCNTs, 0.3 g, 95% purity, Homytech and Conjutek, Taiwan) are dispersed in polyvinyl alcohol (PVA, 2.5 g)-deionized water (80 ml) solution and suspension is ball-milled at 80 °C for 2 h [i, Fig. 2(a)]; (ii) the CNTs/PVA slurry is thermally pressed into A_P-CFCs [300 MPa, 80 °C, ii, Fig. 2(a) and supplementary material, Fig. S1(e)]; (iii) the resultant MWCNTs/PVA-A_P-CFCs composites are heated to 1000 °C in N₂ atmosphere for 1 h [iii, Fig. 2(a) and supplementary material, Fig. S1(f)]. Third, the P_T-CFCs are made by treating H_T-CFCs with O₂/N₂ plasma at gas pressure = 0.2 Torr, ionization power = 150 V, flow rate = 400 c.c/min, and ionization time = 1, 3, and 5 min [Fig. 2(b)]. Plasma treated samples are hereafter defined as P_T(gas)-CFCs_(t), where subscript gas denotes the plasma type and t is ionizing time, e.g., P_T(O2)–CFCs_(1 min) and P_T(N2)–CFCs_(3 min) [supplementary material, Figs. S2(a)–S2(f)]. The samples are analyzed by field-emission scanning electron microscope (FE-SEM, Hitachi SU8010, 15 kV accelerating voltage), Raman (LABRAM HR 800 UV, He/Ne excitation source, λ = 632.8 nm), and x-ray photoelectron emission spectroscopy [XPS, photoelectron instrument (PHI) Quantera scanning XPS microprobe (SXM), Al/Kα radiation source]; the XPS spectra are calibrated with XPSPEAK41 software, NIST database, and C1s peak at 284.6 eV. The S_A is measured and calculated according to nitrogen adsorption at 77 K (NOVA touch 2LX, ASTM D3663-92) and Brunauer–Emmett–Teller (BET) equation. The charge–discharge processes are recorded by cyclic voltammetry (CV) with a three-electrode system in 1M H₂SO₄(aq) (i.e., working, Pt-counter and Ag/AgCl reference electrodes) where the scan window (S_w) and scan rate (S_r) are set at −0.2–0.8 V and 2–100 mV s⁻¹ to avoid electrolysis of water.¹⁵ It is worth mentioning that Cs measured by the three-electrode technique and the equation Cs = (0.5× integral area of CV curve)/(S_r × S_w × m) deviates greatly from true performance and overestimate is due to underestimated mass of active materials (m) and S_w.¹⁶ Accordingly, the Cs, energy, and power density are measured by two-electrode technique at S_r = 100 mV s⁻¹ and L_cyc = 3000, where L_cyc is lifecycle.¹⁷ Electrochemical impedance spectra are recorded at 0.1–1000 Hz and 5 mV, and the obtained Z_re-Z_im profiles (i.e., Nyquist plots) are analyzed with view software to uncover equivalent circuits at interfaces, where Z_re and Z_im represent the real and imaginary parts of diffusive impedance. Ab initio calculation is carried out to characterize the lattice polarity with and without oxygenated groups. First, a graphite sheet consisting of 50 atoms is built and is geometrically optimized with generalized gradient approximation in a 5 × 5 nm superlattice. Second, the ultrasoft pseudopotential and Monkhorst–Pack grid are set at 480 eV and 0.05 A⁻¹, respectively. Third, the Mulliken charge (MC) is used to indicate the degree of lattice polarity. Fourth, the comparison is made on a similar structure with a carbonyl bonded to a hexagonal ring.

Due to the thin coating of lubricants, A_P-CFCs exhibit a hydrophobicity and contact angle reaches 130° [Fig. 3(a)]. H_T removes coatings, and the hydrophilic surfaces emerge at H_T-CFCs, CNTs-CFCs, P_T(O2)–CFCs_(t), and P_T(N2)–CFCs_(t) [Figs. 3(b)–3(d)]. However, the CFC texture does not change with H_T, and the diameter varies only in ±1% (supplementary material, Figs. S1 and S2). Raman spectra of A_P-CFCs and H_T-CFCs show D- and G-bands arising from defects induced phonon scattering at zone boundaries and C–C bond stretching (Table I and Fig. S3).¹⁷ Since the G-band is independent of sheet size and occurs in all sp²-bonded structures, the ratio of D-to G-band intensity (I_D/I_G) can be used to characterize the degree of graphitization, i.e., the lower I_D/I_G the greater degree of graphitization.¹⁷ The measurements give I_D/I_G = 1.0–1.04 for all samples studied, again indicative of the unchanged CFC structure after H_T (Table I and supplementary material, Fig. S3). It is worth mentioning that the conversion of hydrophobicity into hydrophilicity is important since the charge adsorption–desorption requires good wetting and ion accessibility at the electrode surfaces.¹⁸ H_T-improved Cs is therefore anticipated and is verified by Table I extracted from current (I)-S_w profiles [Figs. 4(a)–4(c)], the Cs being measured to be 0.3 F g⁻¹ for A_P-CFCs, 5.75 F g⁻¹ for H_T-CFCs, and 18.9 F g⁻¹ for CNTs-CFCs. Two factors account for CNTs-CFCs_(Cs) > H_T-CFCs_(Cs): (i) the on-tube oxygenated groups act as active sites at the inner Helmholtz layer¹⁹ and (ii) CNT coating creates intertube channels, thus widening S_A.²⁰ Table I supports (ii) and shows a nearly twofold increase. The (i) is evident by redox peaks at 0.58 V for charge and 0.38–0.42 V for discharge [Fig. 4(a)]. Additional evidence in support of (i) comes from P_T promoted Cs and S_A; the maximum reaches Cs = 30.3 F g⁻¹ for P_T(O2)–CFCs_(1 min) and 30 F g⁻¹ for P_T(N2)–CFCs_(5 min) [Table I and Figs. 4(b) and 4(c)].

Oxidation of graphite produces carboxyl, lactol, lactone, anhydride, ether, phenol, carbonyl, quinone, and pyrone.¹⁴ Among others, carbonyl and quinone can act as reactive centers and perform redox reactions in electrolytes [insets, Fig. 4(b)].^14,15 XPS verify the formation of carbonyl, hydroxyl, and carboxyl acids in H_T-CFCs, accounting for conversion of hydrophobicity into hydrophilicity (supplementary material, Fig. S4). Table II lists content of H_T and P_T(O2) created O, C–C, and oxygenated groups according to the integration of the XPS peak area. First, P_T converts the C–C bonds into oxygenated lattices, thus resulting in H_T(O) < P_T(O) and H_T(C–C) > P_T(C–C), i.e., O = 16.46 and C–C = 67.25 at. % for H_T-CFCs and O = 28.07 and C–C = 54.18 at. % for P_T(O2)–CFCs_(1 min). Second, the content of O and C–C changes in the opposite trend (i.e., O increase and C–C decrease) as P_T(O2) is prolonged, consistent with the oxidation mechanism reported by Li and Ross.¹⁶ Third, P_T induced oxygenation is more efficient than does H_T, i.e., P_T(C–O) > H_T(C–O), P_T(C=O) > H_T(C=O), and P_T(O–C=O) > H_T(O–C=O) (Table II). Fourth, the sequence P_T(O2)–CFCs_{(1 min) (Cs)} > P_T(O2)–CFCs_{(3 min) (Cs)} > P_T(O2)–CFCs_{(5 min) (Cs)} is consistent with P_T(O2)–CFCs_{(1 min) (C=O)} > P_T(O2)–CFCs_{(3 min) (C=O)} > P_T(O2)–CFCs_{(5 min) (C=O)}, indicating that FROR operates through the carbonyl groups acting as reactive centers (Tables I and II). The FROR can also proceed through pyrrolic-N in P_T(N2) treated samples [inset, Fig. 4(c) and supplementary material, Fig. S5].^14,15 Elemental analyses suggest that pyrrolic-N probably forms through the reduction of quaternary-N because (i) PAN-based CFCs contain an abundant of quaternary-N and (ii) the content of pyrrolic-N and quaternary-N is complementary (Table III). Tables I and III display a consistency between P_T(N2)–CFCs_(5 min)(Cs) > P_T(N2)–CFCs_(3 min)(Cs) > P_T(N2)–CFCs_(1 min)(Cs) and P_T(N2)–CFCs_{(5 min)(pyrrolic-N)} > P_T(N2)–CFCs_{(3 min)(pyrrolic-N)} > P_T(N2)–CFCs_{(1min)(pyrrolic-N)}, verifying that pyrrolic-N truly acts as reactive centers in FROR.

Question, however, remains how does P_T promote S_A. Three parameters determine S_A according to the BET equation S_A = ϕ_m⋅N_A⋅η/V_m, where N_A, ϕ_m, and V_m denote Avogadro constant, the monolayer uptake, and molar volume of adsorbates. The η is adsorptive area and is of polarity related (i.e., η ∝ S_A).²¹ Figure 5(a) displays an ideal graphite sheet where all atoms are electrically neutral (MC = 0). The introduction of C=O into lattice creates defects where MC becomes −0.38 at O and 0.28 at adjacent C, confirming oxygenation enhanced η [Fig. 5(b)]. Figures 6(a)–6(c) show rate performance, and the data points are extracted from the I-S_w curves (supplementary material, Fig. S6). First, Cs ∝ S_r⁻¹ is present, attributed to time dependent charge–discharge processes.¹ Second, Cs of both P_T(O2)–CFCs_(1 min) and P_T(N2)–CFCs_(5 min) exceeds that of H_T-made samples; the former gives 30 and 20 F g⁻¹ at S_r = 20 and 100 mV s⁻¹, and 30 and 15 F g⁻¹ for the latter. A good retention is observed in P_T samples with L_cyc = 3000 at S_r = 100 mV s⁻¹; Cs being measured to be 83% for P_T(O2)–CFCs_(1 min) and 76% for P_T(N2)–CFCs_(5 min) [Fig. 6(d)].

The equivalent circuits of CFC-based electrodes include a constant phase element (CPE), electrolyte resistance (R_s), and charge transfer barrier (R_ct) connected in series. The CPE is calculated according to Z_CPE = 1/T(iω)^p, where T is CPE constant (unit: Fs^(p–1) cm⁻²), ω is the angular frequency, P is the phase angle, and P = 1 represents an ideal capacitor.²² Figure 7(a) shows Z_re-Z_im profiles of A_p-, H_T-, and CNTs-CFCs, along with the related parameters listed in Table IV. The formation of capacitive reactance due to lubrication coating at A_p-CFCs is again evident by (i) a CPE (semicircle) sandwiched between R_s and R_ct [i.e., a series connection, right insert, Fig. 7(a)], (ii) significant R_s and R_ct, and (iii) CPE-T ∼0 (i.e., a large barrier across a double layer, Table IV).^7,23 The removal of lubrication coating then improves ion admittance and charge transfer, so both R_s and R_ct are greatly reduced (i.e., H_T-CFCs and CNTs-CFCs, Table IV). In this case, the Z_re-Z_im plots become linear at low frequency and R_ct and CPE change to parallel connection [left insert, Fig. 7(a)].²⁴ Similar Z_re-Z_im profiles are also observed in P_T(O2)–CFCs_(t) and P_T(N2)–CFCs_(t), again verifying improved ion admittance and charge transfer at the interface [Figs. 7(b) and 7(c) and Table IV]. The C_s improvement through surface modifications is also supported by increased CPE-P in H_T-CFCs, CNTs-CFCs, and P_T-CFCs_(t) (Table IV).

Figure 8(a) shows a planar capacitor made of two P_T(O2)–CFCs_(1 min) separated by a H₂SO₄ dissolved polyvinyl alcohol film (H₂SO₄ = 1M, 40 ml and PVA, and thickness = 0.1 mm). The experimental setup is shown in Fig. 8(b), where numbers 1, 2, and 3 denote capacitor (weight = 0.135 g), light-emitting diode (LED) (0.09 W), and dry cell (9 V). The capacitor is first connected to dry cell and charged for 10 s. The charged capacitor is then disconnected and rapidly reconnected to LED. We find that LED glows for 60 s and remains dimly lighted between 60 and 90 s (video footage, supplementary material, Fig. S7). Repeated tests give a similar result, and both power density and energy are measured to be 303 W kg⁻¹ and 4.21 Wh kg⁻¹, values which are comparable with FROR governed electrodes.¹ 

In light of data above, it becomes clear that plasma treatment is an important factor in promoting electrochemical performance and dominates Cs improvement over CNT coating (i.e., Cs ∝ ε > Cs ∝ S_A). Question, however, remains as to why P_T(O2)-CFC_(1 min) is superior compared with P_T(O2)-CFC_(3 min) and P_T(O2)-CFC_(5 min). Since the XPS spectra display an opposite trend between the content of O and C–C bonds as plasma treatment is prolonged (Table II), the carbon networks may be severely damaged in P_T(O2)-CFC_(3 min) and P_T(O2)-CFC_(5 min), thus reducing ε and Cs.²⁵ 

The specific capacitance, rate performance, retention, and lifecycle of as-purchased, heat treated, CNTs-coated and plasma treated carbon fiber cloths are studied and compared. CNT coating and plasma treatment significantly improve specific capacitance of carbon fiber cloths made electrodes; the latter clearly dominates improvement over carbon nanotube coating and gives a greater rate performance, retention, and lifecycle. A planar capacitor made from O₂ plasma treated samples and dielectric PVA/H₂SO₄ produces power density = 303 W kg⁻¹ and energy = 4.21 Wh kg⁻¹_.

The supplementary material contains SEM images, Raman spectra, and XPS spectra of as-purchased, heat treated, CNTs-coated and plasma treated carbon fiber cloths. The cyclic voltammetry profiles of CNTs-coated and plasma treated carbon fiber cloths at different scan rates are also shown in the supplementary material. The supplementary video shows that LED glows by charged capacitor that is made of two O₂ plasma treated carbon fiber cloths separated by a H₂SO₄/PVA film.

The authors thank the Ministry of Science and Technology of Taiwan for the financial support (Grant No. MOST- 110-2112-M-007-014). This work was also supported by High Entropy Materials Center from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan (Grant No. 108QR001J4). The authors also acknowledge the use of F200 HRTEM and ARM200 Cs TEM, which belong to Instrumentation Center at NTHU under funding by the Ministry of Science and Technology (MOST), Taiwan.

The authors have no conflicts to disclose.

Hsin-Jung Tsai: Data curation (equal); Investigation (equal); Writing – original draft (equal). I-Hsuan Liao: Data curation (equal); Investigation (equal). Yung-Kai Yang: Writing – original draft (equal). Ding-Jyun Huang: Data curation (equal). Wen-Kuang Hsu: Project administration (equal); Writing – review & editing (equal).

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