The YP and SG were obtained from Genera Energy LLC (Vonore, TN). Lignin was isolated using an γ-valerolactone organosolv fractionation technology described by Alonso et al. (2017) at the lab scale. YP and SG lignin were carbonized in a ceramic boat inserted in an alumina tube furnace with nitrogen gas flowing (3 L min⁻¹) at 1,000°C for 1 h. After naturally cooling down with N₂ flow, the carbonized samples were pulverized by ball-milling (PM100 RETSCH model), using 2- and 10-mm stainless steel balls in a stainless-steel container at 350 rpm for 30 min. The grounded carbon powders were separated from the grinding balls using sieves and then placed into the tube furnace for steam activation at 800°C for 1 h with a ramping rate of 10°C min⁻¹. After steam activation, the carbon sample was placed in the furnace under CO₂ atmosphere for physical activation at 800°C for 1 h to yield the final samples. Figure 1 illustrates the schematic diagram for the experimental design for the supercapacitor using two kinds of electrode materials (i.e., YP and SG), where activated carbons (YP-AC and SG-AC) were derived from different raw materials and lignin precursors.

Thermogravimetric analyzer (TGA, Perkin Elmer Pyris 1) and differential scanning calorimetry (DSC, Perkin Elmer Diamond DSC) were adopted to measure thermal stability of lignin precursors in an inert nitrogen environment. The TGA analysis was conducted by heating lignin samples from 100 to 800°C at a heating rate of 10°C min⁻¹. The DSC procedure involved three heating and cooling cycles performed at a rate of 100°C min⁻¹. The lignin samples were initially heated for two cycles from 25 to 160°C, and then heated to 230°C.

The structural morphology images of the carbon samples were collected by using scanning electron microscope (SEM, Phenom ProX). Surface characteristics of the carbon samples were analyzed by an automated adsorption apparatus (Micromeritics Instrument Co., TriStar 3000) using N₂ physisorption at -196°C. Total and mesopore surface areas of the carbon samples were calculated by using Brunauer–Emmett–Teller (BET) and Barrett–Joyner–Halenda (BJH) equations. Pore-size distributions were analyzed from the N₂-adsorption branch using density functional theory (DFT) method.

Surface functional groups on the activated carbon samples were characterized by attenuated total reflectance Fourier transform infrared (FTIR) spectroscopy (Perkin Elmer, Spectrum II). FTIR spectra were collected from 4,000—600 cm⁻¹ using 16 scans with a resolution of 4 cm⁻¹. Additional surface chemical composition was collected by X-ray photoelectron spectroscopy (XPS, Fison VG ESCA210). Elemental (i.e., CHN) analysis of lignin samples was performed by combustion analysis (Perkin Elmer) to determine amounts of carbon, hydrogen, and nitrogen. To quantify the amount of oxygen, a differential technique was adopted where the samples were assumed to be composed of solely C, H, N, and O elements. Gel permeation chromatography (GPC) was used to determine the number average molecular weight (Mn), weight average molecular weight (Mw), and molar-mass dispersity (Ð _M) index of the lignin samples. GPC analysis was conducted by dissolving lignin in tetrahydrofuran (THF) at 1 mg ml⁻¹ and injecting into an EcoSec GPC Instrument (Tosoh), with two columns (TSKgel SuperMultipore HZ-M, 4.6 × 150 mm, 4 µm) and a guard column held at 40°C. Acetobromination of the lignin was required prior to dissolution in THF for GPC analysis (Guerra et al., 2008).

³¹P-NMR measurements were carried out using a Varian 400-MR spectrometer. Lignin samples, approximately 60 mg, were dissolved in 0.4 ml NMR solvent (deuterated chloroform and pyridine (1:1.6 v/v)). Then, 0.1 ml internal standard solution (21.5 mg endo-N-hydroxy-5-norbornene-2,3-dicarboximide in 1 ml NMR solvent) and 0.05 ml relaxation reagent (11.4 mg chromium (III) acetylacetonate in 1 ml NMR solvent) were added. Immediately before NMR analysis, 0.1 ml phosphorylating reagent, 2-chloro-4,4,5,5-tetramethyl-1,3,2- dioxaphospholane (Santa Cruz Biotechnology) was added to the sample. Acquisition parameters were set to collect 512 scans, with 1 s acquisition time and 5 s relaxation delay. Spectral data were analyzed using MestReNova software.

Two-dimensional (2D) ¹H ‒¹³C NMR spectra were conducted on a Bruker Avance III 500-Hz NMR spectrometer, and spectral processing was carried out using a Bruker Topspin 3.5 software. The lignin samples (30 mg) were dissolved in 0.4 ml DMSO-d ₆ in an NMR tube independently. Heteronuclear single quantum coherence (HSQC) experiments were perfomed with a Bruker pulse sequence (hsqcetgpspsi 2.2) on an N₂ cryoprobe (BBO ¹H&¹⁹F-5mm) with the following acquisition parameters: spectra width 12 ppm in F2 (¹H) dimension with 1,024 data points (acquisition time 85.2 ms), 200 ppm in F1 (¹³C) dimension with 256 increments (acquisition time 5.1 ms), a 1.0-s delay, a ¹ J _C–H of 145 Hz, and 128 scans. The central DMSO-d ₆ solvent peak (™_C/™_H at 39.5/2.49) was used to calibrate the chemical shift position. Assignment and the relative abundance of lignin compositional subunits and inter-unit linkage were estimated using volume integration of contours in HSQC spectra according to published literature (Das et al., 2018). The abundances of aromatic units and side-chain linkages were presented as percentage of total SGH units and total side-chain linkages, respectively.

Prior to working electrode preparation, the electrode slurries were first prepared by mixing 94.0 wt% activated carbon powder, 5.0 wt% polyvinylidene fluoride (Kynar) and 1.0 wt% black carbon (Super C65) in N-methyl-2-pyrrolidone (NMP) solvent. Subsequently, the stainless-steel substrate was coated with the electrode slurry to obtain a flexible electrode with an area of 2 × 1 cm².

Electrochemical measurements were conducted in a three-electrode system with a Pt wire as the counter electrode and an Ag/AgCl electrode as the reference electrode. Sulfuric acid (H₂SO₄, 1 M) was used as the electrolyte solution. The CV measurements were conducted at different scan rates (10, 30, 50, and 100 mV s⁻¹) within the potential range of 0–1 V. The capacitance was measured by charging and discharging from 1 to 0 V at different current densities. The impedance spectra were recorded with the AC frequency ranging from 100 kHz to 1 mHz. The resulting Nyquist plots were analyzed using the Z-view software package.

The compositional and elemental analysis of various lignin precursors are provided in the Electronic Supporting Information (see Supplementary Table S2). Supplementary Table S2 includes the purity analysis, average molecular weight, and elemental analysis. Supplementary Table S2 details that there is less than 4 and 6% of contaminants for YP and SG, respectively. Additionally, the ash contents are low, 0.02 and 0.03% for YP and SG, respectively, indicating that YP lignin has higher purity than SG lignin. GPC analysis details that YP lignin possesses similar molar-mass dispersity (Ð _M) index (M _w/M _n: 3.00 (YP) and 2.96 (SG)) with SG lignin, which reveals the dispersion of the distribution of molar masses in YP and SG lignin (Kilpelaeinen et al., 1994; Kubo and Kadla, 2005). The quantitative CHN of these two lignin sources are summarized in the elemental analysis, showing that YP lignin possesses a higher carbon content than SG, whereas SG lignin has higher oxygen and nitrogen content.

Hydroxyl units are a key factor that affect the product’s thermal and physical properties, which was quantified using ³¹P NMR (Hosseinaei et al., 2016; Hosseinaei et al., 2017). As shown in Figure 2 and Table 1, YP and SG have similar amount of aliphatic hydroxyl groups, while SG possess a higher amount of phenolic hydroxy groups and carboxylic acid groups. The higher level of carboxylic acid in SG lignin is likely a result of higher concentration of phenolic acids (pCA and FA) being present in grasses (Hosseinaei et al., 2016). Since YP lignin predominantly contains guaiacol (G) units, the amount of p-hydroxyphenyl OH groups in SG lignin are higher than YP. To further confirm above results, we used 2D NMR analysis quantify the aromatic and aliphatic region in both lignin types, as illustrated in Figure 3. Supplementary Table S3 includes the relative abundance of lignin linkages assessed using ¹³C-¹H NMR analysis. The results agree with ³¹P NMR results confirming higher degree of lignin condensation in YP relative to SG due to the presence of the C₅ position of G units, which provide coupling and cross-linking sites. The linkages of methoxyl (OMe), γ-hydroxylated (A _γ ), β-5′ phenylcoumaran (B _γ and B _β ) groups and β- O-4’ (A _α /A _α ’) can be observed for both lignins. However, SG lignin contains resinol (C _β and C _γ ) substructures, which YP lignin does not possess.

Lignin’s chemical composition depends on the botanical source and isolation method. Lignin from grasses (e.g., SG) usually possesses all three units (H-, G-, and S-type), whereas softwood lignin (e.g., YP) are primarily composed of G-type units (Yoshida et al., 2004). As shown in Figure 3 S– and G-units and H-units are clearly observed in the aromatic region of the SG sample, while YP predominantly possess G-units along with a small quantity of H-units. The S-units show strong signals for C₂/H₂ and C₆/H₆ in correlation with δC/δH 103.9/6.65 (S_2/6), the G-units display strong correlations at δC/δH: 110.9/7.00 (G₂), 115.1/6.72and6.98 (G₅), and 118.7/6.77 (G₆), and the peaks of H-units are observed at 114.8/6.73 (H_3,5) and 127.8/7.20 (H_2,6).

Aromaticity, an indicator of a product’s quality as well as its thermal stability, was evaluated for the precursors using TGA, as shown in Figure 4A. The pyrolysis of both lignin types presents a similar trend of weight loss starting at 200°C, which corresponds to the breaking of the weaker bonds such as hydrogen bonds and C-OH binding (Liu et al., 2015). Decomposition is observed in the range of 200–300°C, which proceeded faster for SG due to its higher number of linkages. The maximal loss occurs at 380–390°C, where the cleavage of stronger bonds, such as β-O-4 linkages begins to degrade (Liu et al., 2015). The maximal decomposition temperature of YP is higher than that of SG, at 395 and 385°C respectively. This implies that YP lignin is more thermally stable compared to SG lignin with the presence of a more condensed molecular structure. Eventually, a carbon residue forms at 700°C with approximately 40 wt% fixed carbon. Figure 4B presents typical DSC curves of both lignin structures, which is used to determine the glass transition behaviors of the lignins. The glass transition temperature (T _g) of SG lignin (155°C) is higher than that of YP lignin (142°C) due to more hydroxyl groups in SG samples. Hydroxyl groups play an important role in affecting the thermal properties of lignin, especially the T _g. The existence of hydroxyl groups leads to the formation of hydrogen bonds that subsequently limits molecular movement (Kubo and Kadla, 2005; Hosseinaei et al., 2017).

Figures 5A,B demonstrate SEM images of pyrolyzed lignin from YP and SG lignin (i.e., YP-py and SG-py) obtained upon thermal pyrolysis at 1,000°C. As illustrated in Figure 5, the pyrolyzed lignin exhibits contiguous morphology where the YP-py appears as flat sheets containing numerous cavities, while the SG-py is formed as curved structure. This observation reflects that lignin carbonization occurs during the pyrolysis process, inducing decomposition along with char formation. This dynamic interaction results in releasing gaseous compounds, including oxygen and hydrogen in the form of CO, CO₂, H₂, CH₄, etc. (Rodríguez Correa et al., 2017). This promotes the formation of condensable volatiles due to free radicals generated after the cleavage of inner bonds, usually the C‒O bond in the β- O-4’ substructures (Britt, 1995). Of note, S-type lignin includes a higher amount of C‒O bonds compared to H- and G-type lignin. The quantitative CHN comparison between these two lignin structures also confirms this trend (i.e., the SG lignin contains higher O/C ratio (∼50.4%) than YP (∼44.3%), as summarized in Supplementary Table S2). This correlation implies the SG lignin is more easily decomposed compared to the YP lignin upon thermal pyrolysis. After pyrolysis, the lignin is converted into char with high carbon content and abundant surface functional groups, but limited surface area (Liu et al., 2015). As shown in Supplementary Figure S1, XPS survey-scan spectra confirms the presence of carbon (C 1s) and oxygen (O 1s) in both pyrolyzed lignin samples. In addition, it is notable that there is a small amount of inorganic impurity, which is sodium (Na 1s) peak located at ∼1,072 eV, in YP-py sample. Elemental analysis shows the O/C ratio of SG-py (∼8.1%) is lower than YP-py (∼15.2%), and there is ∼3.6% Na in YP-py. The surface area obtained by BET measurements of YP-py and SG-py samples are all negligible (YP-py: 0.02 m² g⁻¹ and SG-py: 0.11 m² g⁻¹). It is well documented that both surface area and pore volume increase as a result of the activation process (C.-T, 1998). The increase in porosity of YP-AC and SG-AC can be attributed to the opening of closed pores and/or enlarging of the micropores due to gasification of the carbon during the activation process (Kinoshita, 1988).

SEM images of activated carbons prepared from YP and SG lignin are illustrated in Figures 5C,D, respectively. According to Figure 5, it is obvious that the smaller carbonaceous aggregate size, and greater porosity is present with YP-AC compared with SG-AC. Supplementary Table S4 includes characteristics of different porous carbons determined from analysis of the N₂ adsorption isotherms at -196°C. Based on BET measurements, the YP-AC sample possesses a specific surface area of ∼1,070 m² g⁻¹, which is approximately two times higher than that of SG-AC (∼507 m² g⁻¹). The YP-AC sample contains a higher percentage of mesopore (∼44%), whereas the SG-AC is mainly microporous due to the presence of 89% micropore volume. The DFT method was employed to analyze the full-range pore size distribution of both carbon samples, as shown in Supplementary Figure S2. The DFT analysis reveals different pore size distributions within the structure, where the YG-AC sample possesses a sharp peak at the pore size of ca. 1‒2 nm, and a smaller but broad distribution ranging up to 50 nm, while the SG-AC sample exhibits one major lump at ca. 1–2 nm. Since both carbon samples were thermally treated through a similar pyrolysis and two-step activation method, this variation in the morphology of the activated porous electrodes implies that surface area, porosity, and pore size distribution is strongly dependent on the lignin nature.

Physical activation usually involves three mechanisms, including surface burn-off, pore deepening, and pore widening during the carbon gasification (i.e., activation process). For the YP carbon with high O/C ratio, it is more prone to surface burn off and the oxidation reaction induced by the oxygen functional groups that facilitate the pore widening by destroying the microporous structures (Teng and Hsieh, 1998). On the other hand, the extent of carbon agglomeration in YP-py is less than that in SG-py, indicating more surface area of YP-py is accessible to CO₂/H₂O upon activation. More pores could be evenly widened and the closed micropores are allowed to be opened during gasification (Teng et al., 1996). In addition, it has been demonstrated that the presence of Na has a significant catalytic effect during the activation, (Rodrguez-Mirasol et al., 1993; Liu et al., 2015), which contributes to the high surface area of YP-AC. As for the SG carbon, with relatively low O/C ratio, the accessibility of the activation agents, H₂O and CO₂, increases with the extent of the burn-off and subsequently results in a parallel reaction step with pore deepening and pore widening. The difference between the YP and SG chars results in different porosities as well as pore size distributions. Therefore, the O/C ratio, char structure, and inorganic impurities are key factors that lead to highly porous ACs.

Elemental analysis shows that the YP-AC and SG-AC contain high O/C atomic ratios of 7.3 and 10.5%, respectively. This observation reveals that both produced ACs possess high oxidation levels (i.e., many oxygen functionalities attached to porous carbon). XPS spectra (see Figure 6A,B) were obtained to confirm the presence of surface functional groups on the carbons. The full spectra are presented in Figure 6A with the typical C 1s and O 1s peaks located at ∼285 and ∼531 eV, respectively. The C 1s peak (shown in Figure 6B) was split into three peaks located at 284.6 ± 0.1 eV (C=C or C–C), 286.5 ± 0.1 eV (C-O) and 288.9 ± 0.1 eV (C=O or O-C=O). This result is further confirmed by FTIR spectra, which are displayed in Figure 6C,D. The peak occurring at ca. 1,540 cm⁻¹ is assigned to the presence of C=C stretching (Tang et al., 2012). The strong peaks in the range of 800–1,300 cm⁻¹ are due to stretching of the C‒O and–COOH bonds (Yang et al., 2012; Russo et al., 2014; Vinayan et al., 2016). The transmittance peak appears at around 1740 cm⁻¹, mainly originating from the C=O stretch in various functional groups (i.e. ketones, carbonyls, esters, carboxylic acid groups, etc.) (Periasamy et al., 2009; Kharangarh et al., 2018). Based on the FTIR spectra, oxygen functional groups exist on both activated carbons samples, attached to edge or basal plane of carbons.

Typical CV curves of capacitors assembled with YP and SG carbons were recorded within the potential range of 0–1 V at various sweep rates as illustrated in Figure 7A,B. As shown in Figure 7, the voltammograms of the carbon electrodes display an obvious pair of redox peaks at 0.3–0.5 V. The CV profiles demonstrate a pseudo capacitance behavior where the induced current increases with increasing sweep rate. Additionally, a rectangular voltammogram, in which the current quickly reaches and maintains a relatively constant magnitude (Figure 7) upon reversal of the potential sweep can be maintained, although it shows current leakage at such high sweep rates (e.g., 100 mV s⁻¹) for both capacitors. The current density of the supercapacitors assembled with the YP carbon electrode was higher compared to that of SG electrodes due to larger available surface area of the YP carbon structure, enabling 1) the formation of electric double-layer and 2) the redox reaction in the presence of oxygen functionalities.

The formation of carbon electric double layer with acidic electrolyte and subsequent charge/discharge processes can be formulated using equation (R1) (Zheng et al., 1997; Hsieh and Teng, 2002): C   ∗ +   H + ↔ C ∗ / / H + (R1)

Here, C* is the active carbon sites, H⁺ represents the protons, and the double layer where the charges accumulated separately on two sides of the interface is shown as//. It formulates a physical adsorption process induced by electrostatic forces between the carbon substrate and the protons. The pseudocapacitance, originated from the oxygen functional groups with electron-accepting properties, such as carbonyl and quinone groups, also induces redox reactions during the electron transfer process (Hsieh and Teng, 2002; Conway, 2013): C x O ∗ + H + + e − ↔ C x O H (R2) where C _x O* represents carbonyl and quinone groups. Since YP carbon has higher porosity with an increased oxidation level, it provides more hydrophilic sites accessible for the double-layer formation and redox reaction, promoting the progress of (R1) and (R2) reactions.

Figure 8A,B show typical charge-discharge profiles of YP and SG capacitors at 0.5 and 2.5 mA cm⁻², respectively. The YP capacitor exhibits higher capacitance for both charge and discharge cycles and demonstrates reduced ‘IR drop’ at the beginning of discharge. Such a small “IR drop” is mostly due to higher ionic conductivity and lower diffusion resistance of the electrolyte within the porous carbon structure (Conway, 2013). Also, as shown in Figure 8, one charge-discharge plateau at 0.4–0.5 V, attributed to the pseudocapacitance from the redox reaction step (R2). The variation of specific capacitance with discharge current density, from 0.1 to 2.5 mA cm⁻², is depicted in Figure 8C. At 0.1 mA cm⁻², the specific capacitances of YP and SG were ∼367 and 221 F g⁻¹, respectively. The specific capacitance was found to decrease with increasing discharge current, which displays a trend confirmed by other research (Hsieh et al., 2010).

To assess the stability of as-prepared porous electrodes, extended charge-discharge cycling was performed at 10 mA cm⁻². The capacitors equipped with YP-AC and SG-AC electrodes were charged and discharged between 0 and 1 V repeatedly, as shown in Figure 8D. Analyzing the variation of discharge capacitance with cycle number revealed that the YP capacitor has superior and stable capacitance (capacitance retention >90%) and exhibits excellent Coulombic efficiency (>99%) over long-term cycling (>10,000 cycles). The weakening of the Faraday reactions after long-term cycling is one of the reasons that cause the slight decrease of capacitance retention (∼94%) after 10,000 cycles. We also noticed that both capacitors show the trend during cycling: the capacitance first decreases then increases. We speculate that this phenomenon results from the electrodes needing time to fully wet the complex porous structures, especially at fast charge and discharge rates. YP-AC electrode took a longer time than SG-AC, as the surface area of YP-AC is doubled that of SG-AC.

The EIS technique was employed to analyze the electrochemical behavior of the capacitors assembled with various lignin-based electrodes. The impedance spectra of different capacitors are shown as Nyquist plots in Figure 9A. As illustrated in Figure 9, the high-frequency intersection with the real axis represents the electrolyte resistance (R _E), followed by a single quasi-semicircle in the high-frequency region (100 kHz - 1 Hz). The semicircle in the high-frequency region is due to combined effect of interfacial resistances (between activated carbon composites and current collector (R _I//C _I)), as well as the charge transfer impedance (R _CT//C _DL, a double-layer capacitance in parallel with a resistance) (Kinoshita, 1988). The EIS spectra also include a linear regime in the low frequency region corresponding to the capacitive response of the porous carbons (McCreery et al., 1994). The equivalent circuit model used for simulating the impedance behavior of the capacitors is shown in the inset of Figure 9A. The Z-view software package was adopted for fitting the EIS spectra with the equivalent circuit model. The overall resistance, so-called “equivalent series resistance”, R _ES (= R _E + R _I + R _CT), was composed of the electrolyte resistance, interfacial impedance, and charge transfer resistance. The R _ES values were 28.7 and 24.8 Ω for YP and SG capacitors, respectively. The slightly higher R _ES value of YP electrode is attributed to higher surface area and more complex porous structure, resulting in the high resistance for charge transfer within the pore structure.

As shown in Figure 9A, the Nyquist plot displays an inclined line at low frequencies (<1 Hz), resulting from the solid-state diffusion process of protons within the porous electrodes. Meanwhile, Warburg impedance (Z _D) in the electrodes can be formulated as Z _D = k _W ω ^−1/2 (1 − j), based on semi-infinite diffusion model. Here, k _W represents the Warburg coefficient and ω (= 2πƒ) is the angular frequency (f). Figure 9B presents the “Randles plot”, Re Z’ versus ω ^−1/2, for both capacitors. The diffusion coefficients (D) determined from the k _W value were 4.77 × 10⁻¹⁰ and 2.16 × 10⁻¹⁰ cm² s⁻¹ for the YP and SG capacitors, respectively, detailing the D value of YP capacitor was more than two times higher than that of SG. This finding reveals that the ionic diffusion resistance is significantly alleviated for the YP electrodes, leading to fast redox reaction (i.e., R2) and rapid formation of electric double-layer (i.e., R1) in the mesoporous carbon. Accordingly, the unique pore texture of the YP structure significantly facilitates the electrochemical reactions. Shorter diffusion path, along with increased mesopores and surface area available for energy storage with the YP carbon structure, were the major contributing factors for the improved performance.

To demonstrate the economic feasibility of the approach, the minimum selling price (MSP) of activated carbons produced by using YP and SG as feedstocks is estimated. Four steps are taken to perform techno-economic analysis.