Thermogravimetric kinetic analysis of mallow (Urena lobata Linn.) stem residues to evaluate their potential in energy applications

1 Introduction

Society has been under increasing pressure to change its fossil fuel consumption patterns due to growing concerns about environmental impacts. The intensive use of coal, oil, and natural gas is directly associated with greenhouse gas emissions and the exacerbation of global warming (Regueiro-Ferreira and Alonso-Fernández, 2023). Consequently, the search for renewable and regionally available energy sources has intensified, with biomass standing out due to its availability, renewable nature, and potential contribution to energy security (Rehman et al., 2019; Araujo et al., 2022).

Biomass can be defined as the biodegradable fraction of products, residues, and materials of biological origin derived from agricultural, forestry, and industrial activities (Zhang, 2025). Its composition varies according to plant species and tissue type, mainly in terms of hemicellulose, cellulose, lignin, extractives, and inorganic matter, which directly influence its thermal behavior during pyrolysis (Afzal et al., 2025; Kuo et al., 2023; Li and Paul, 2025). The thermal degradation of lignocellulosic biomass is strongly associated with the structure and stability of its main components.

The mallow (Urena lobata Linn.), a fibrous species belonging to the Malvaceae family and native to the Amazon region, has an estimated annual production of 4.500 tons, according to data from the Brazilian Institute of Geography and Statistics (IBGE). Its main commercial product is the fiber, which is separated from the stem through maceration and used in the textile industry, mainly in the manufacture of sacks, fabrics, and rugs. After fiber extraction, the remaining stem material represents a significant lignocellulosic byproduct that is often discarded or underutilized. This situation creates an environmental burden and represents a loss of material with potential for reuse. Thus, the valorization of mallow stem residue may contribute to reducing waste accumulation, diversifying applications, and adding value to a traditional amazonian production chain without compromising the primary product, the fiber.

Thermochemical conversion technologies, including gasification, pyrolysis, torrefaction, liquefaction, and hydrothermal carbonization, have been widely investigated for transforming biomass into energy products and value-added materials. In this context, kinetic analysis is essential for understanding decomposition rates, product distribution, and the formation of biochar, bio-oil, and gases. Kinetic modeling serves as a bridge between thermoanalytical characterization and process engineering, since it allows the quantification of kinetic parameters and the representation of global or multistage mechanisms involved in biomass conversion (Kumar et al., 2025).

Advanced isoconversional methods have been widely applied to capture variations in reactivity and to describe complex degradation processes as a function of conversion (Kumar et al., 2025; Yang et al., 2024). In isoconversional methods, kinetic parameters are estimated without assuming a single reaction model f(α). The activation energy as a function of conversion, E _*a* (α), is obtained using multiple heating rates, including differential methods such as Friedman, integral methods such as FWO, KAS, and Starink, and numerical formulations such as Vyazovkin’s nonlinear method (Gupta et al., 2025; Yiga et al., 2025). These approaches are robust for complex systems because they capture changes in reaction mechanisms throughout the decomposition process and help evaluate the consistency of experimental data (Ferfari et al., 2025). However, they do not directly provide a unique reaction model and may be influenced by derivative noise, conversion range selection, and heat and mass transfer limitations, which justifies the combination of different methods and complementary interpretations (Mariyam et al., 2023; Zou et al., 2022). The use of KAS, FWO, and Starink has been reported for several biomasses, including corn stalk, mangrove biomass, sawdust, oat straw, bamboo, and flax shives, with activation energy values varying according to feedstock composition and conversion range (Paredes et al., 2024; Khalid et al., 2024; Hossain et al., 2025; Jalilian et al., 2025).

Therefore, this study aims to perform a thermogravimetric kinetic analysis of residual biomass from mallow (Urena lobata Linn.) stems to evaluate its energy potential for power generations purposes. For this purpose, the FWO, KAS, Starink, and Vyazovkin isoconversional methods were applied to estimate conversion-dependent activation energy, while thermodynamic parameters and reaction mechanism assessment were used to provide a deeper understanding of the energetic requirements and degradation pathway of this underutilized amazonian biomass residue.

2 Experimental

2.1 Biomass preparation and characterization

Mallow (U. lobata Linn.) was harvested at the experimental farm of the Federal University of Amazonas (UFAM), located along BR-174 Highway, Km 38, Manaus, Amazonas, Brazil (2°38′59.8″S; 60°03′11.6″W). After fiber extraction, stem samples were cut and oven-dried at 105 °C for 5 h and subsequently ground in a knife mill and sieved to a particle size of 60 mesh. The lignocellulose, extractives, holocellulose, and lignin contents of the biomass samples were determined in accordance with the procedures of the Technical Association of the Pulp and Paper Industry (TAPPI), TAPPI T-204 CM-97 and TAPPI T-222 OM-02. All analyses were performed in triplicate.

The chemical composition analyses, including extractives, holocellulose, α-cellulose, hemicellulose, and lignin contents, were performed in triplicate and expressed as mean ± standard deviation (n = 3). Hemicellulose content was calculated from the difference between holocellulose and α-cellulose contents. These values were used as experimental reference data for comparison with the apparent lignocellulosic fractions estimated by DTG deconvolution.

The proximate parameters used for the estimation of the higher heating value were obtained from the TG data under an inert N₂ atmosphere. Therefore, the calorific value reported in this study should be interpreted as an estimated higher heating value (HHV _*est*) calculated from an empirical correlation, and not as an experimental value measured by bomb calorimetry. The HHV _*est* was calculated using the correlation proposed by Parikh et al. (Parikh et al., 2005), as shown in Equation 1:where FC, VM, and ASH correspond to fixed carbon, volatile matter, and ash contents, respectively, expressed as mass percentages. In this study, FC and VM were estimated from the TG/DTG mass-loss events as 5.428% and 72.555%, respectively. After revision of the TG mass balance, the final residue at 700 °C under N₂ atmosphere was corrected to 5.570%, corresponding to approximately 0.274 mg from an initial sample mass of 4.919 mg. This corrected value was used as an approximate residual fraction in the HHV estimation.

H H V e s t = 0.3536 F C + 0.1559 V M − 0.0078 A S H (1)

2.2 Thermogravimetric Analysis and TG-MS

Thermogravimetric analyses were conducted using a Shimadzu TG-60H thermogravimetric analyzer, employing approximately 5.0 ± 0.5 mg of sample placed in an alumina crucible. The experiments were performed under an inert nitrogen (N₂) atmosphere, with a controlled flow rate of 20 mL min^-1, over the temperature range of 25 °C–700 °C, at heating rates of 2.5, 5, 10, 15, and 20 °C min^-1. The thermogravimetric analyses were performed using a Shimadzu TG-60H thermogravimetric analyzer, employing approximately 5.0 ± 0.5 mg of sample placed in an alumina crucible. The experiments were conducted under an inert nitrogen (N₂) atmosphere, with a controlled flow rate of 20 mL min^-1, in the temperature range of 25 °C–700 °C, at heating rates of 2.5, 5, 10, 15, and 20 °C min^-1. Prior to the experiments, the instrument was stabilized under the same gas flow conditions used in the analyses. Baseline correction and processing of the TG curve data were performed using Shimadzu’s thermogravimetric analysis software prior to data export. No additional correction was applied after processing. However, potential baseline drift was assessed by verifying the initial and final regions of the TG curves. The corrected TG data were then used to obtain the corresponding DTG curves using the same Shimadzu software, ensuring consistency between the TG mass loss and the derivative thermogravimetric profile.

The gases evolved during the pyrolysis experiments were analyzed using thermogravimetry–mass spectrometry (TG–MS), employing a NETZSCH QMS 403 °C Aeolos® mass spectrometer coupled to a STA 449 F1 Jupiter® thermal analyzer (NETZSCH). Gas transfer was performed using a 75-μm-diameter quartz capillary tube enclosed in a stainless-steel sheath. To prevent condensation of volatile products, the transfer line was maintained at 270 °C. Gas ionization was achieved by electron impact at 70 eV. The experiments were conducted under an inert argon atmosphere, used as the carrier gas at a flow rate of 20 mL min^-1, with a heating rate of 10 °C min^-1. The qualitative identification of the main volatile products was based on the intensities of the mass-to-charge ratios (m/z), with emphasis on the signals at m/z 12, 15, 18, 30, 44, 46, 53, 55, 57, 60, 64, 74, and 81.

2.3 Kinetic parameters of biomass

The thermal decomposition of lignocellulosic biomass involves a series of multistage chemical transformations. Understanding the reaction mechanism requires determining the so-called kinetic triplet, consisting of the activation energy (E _*a*), the pre-exponential factor (A), and the reaction model f(α). During thermal degradation, biomass comprising different chemical components results in the formation of non-condensable gases, bio-oils, and char (Santos et al., 2023). The progression of the reaction can be described mathematically using the degree of conversion, as shown in Equation 2:

α = m 0 − m t m 0 − m f (2)

Where m _*0*, m _*f*, and m _*t* are the initial mass, final mass, and instantaneous mass of the biomass during TG analysis, respectively. For kinetic calculations, m _*0* and m _*f* were defined within the main pyrolytic degradation region, excluding the initial moisture/light volatile release stage and the final carbonaceous residue stabilization stage. The kinetic analysis was performed in the conversion range of α = 0.10–0.90, using fixed conversion intervals of Δα = 0.05. This interval was selected because it corresponds to the main mass-loss region of the TG/DTG curves and minimizes the influence of low-conversion moisture removal, baseline instability, and high-conversion secondary char-forming reactions. For each heating rate, the temperature corresponding to each conversion degree, Tα, was obtained from the TG curve by interpolation and used in the FWO, KAS, Starink (STK) and Vyazovkin (VYZ) calculations.

The reaction rate during pyrolysis depends on temperature (T) and the degree of conversion (α), and can be mathematically expressed by Equation 3:where the constant k(T) can be expressed by the Arrhenius Equation 4:

d α d t = k T f α (3)

k T = A e − E a R T (4)

Combining Equations 3, 4, the reaction rate can be expressed according to Equation 5:

d α d t = A e − E a R T f α (5)

Considering the heating rate (β) in Equation 6:The reaction rate can be expressed as a function of temperature, as shown in Equation 7:where A is the pre-exponential factor, E _*a* is the activation energy, R is the universal gas constant, T is the absolute temperature, and f(α) is the differential reaction model. The corresponding integral form is expressed as g(α), according to Equation 8:where p(x) is the temperature integral and x = E a R T. Since p(x) does not have an exact analytical solution, numerical approximations were used, including those proposed by Doyle, Murray and White, Starink and Tang, and the Senum–Yang approximation when required (Mishra and Mohanty, 2020).

β = d T d t = d T d α x d α d t (6)

d α d T = A β e − E a R T f α (7)

g α = ∫ 0 α d α f α = A β ∫ T 0 T e − E a R T d T = A E a β R p x (8)

For the linear isoconversional methods, FWO, KAS, and Starink, activation energy values were obtained from linear regressions performed at each conversion degree using the five heating rates. The quality of the linear fits was evaluated using the coefficient of determination (R ²), while the uncertainty associated with the activation energy was assessed from the regression statistics, particularly the standard error of the slope when applicable. For the Vyazovkin method, numerical stability was evaluated by repeating the optimization procedure using different initial activation energy estimates and verifying convergence to the same minimum of the objective function.

2.4 Flynn-Wall-Ozawa (FWO) method

The Flynn–Wall–Ozawa (FWO) method is a linear integral approach that employs the approximations proposed by Doyle for determining the activation energy (E) (Ozawa, 1965). Under these conditions, the relationship can be expressed by Equation 9:

ln ⁡ β = ln A E a R g α − 5.331 − 1.052 E a R T (9)

Where the activation energy (E _*a*) is obtained from the slope of the line – 1.052 · (E _*a* /R), determined from the graph of ln(β) as a function of the inverse of temperature (1/T).

2.5 Kissinger-Akahira-Sunose (KAS) method

To determine Ea using the KAS method, the calculation is performed according to the final equation, Equation 10:

ln β T 2 = ln R A E a g α − E a R T (10)

Where E _*a* is obtained from the slope of the line equal to − E a R from the graph of ln β / T 2 versus the inverse of temperature (1/T) at different heating rates β.

2.6 Starink’s method

Using the Starink method (Starink, 2003), the activation energy is obtained in a manner similar to that of FWO and KAS, as shown in Equation 11:

ln β T 1.92 = C o n s t a n t − 1.008 E a R T (11)

Where E _*a* was estimated from the slope of the line, which is equal to − 1.008 E a R obtained from the graph ln β / T 1.92 the inverse of the temperature (1/T).

2.7 Vyazovkin method

The Vyazovkin method is a nonlinear isoconversional approach that allows the determination of conversion-dependent activation energy values using numerical minimization (Vyazovkin and Dollimore, 1996). In this study, the apparent activation energy was obtained by minimizing the objective function Ω (Equation 12) individually at each conversion level within the range α = 0.10–0.90.

Ω = min ∑ i ≠ 1 k ∑ j ≠ i k β j l E a , T α . i β i l E a , T α . i (12)

The five heating rates, β = 2.5, 5, 10, 15, and 20 °C min^-1, were used in the optimization procedure. The temperature integral I (E _*a* ,T _*α*) was calculated using the 8th-order Senum–Yang approximation (Araujo et al., 2022; Senum and Yang, 1977). The minimization was performed using the generalized reduced gradient algorithm available in Microsoft Excel Solver, version 2,105. The activation energy was searched within the range of 50–250 kJ mol^-1, using an initial estimate of 120 kJ mol^-1. To verify numerical stability, the optimization was repeated using different initial estimates, and convergence was accepted when changes in the objective function were lower than 10⁻⁶ and variations in E _*a* between consecutive iterations were lower than 0.01 kJ mol^-1. Solutions that converged to the same minimum from different initial estimates were considered numerically stable.

2.8 Thermodynamic parameters–conventional approach

The pre-exponential factor (A), enthalpy change (ΔH), Gibbs free energy change (ΔG), and entropy change (ΔS) were calculated using Equations 13‐16, respectively, as follows:

Δ H = E a − R T (13)

Δ G = E a + R . T m . ⁡ ln K B . T m h . A (14)

Δ S = Δ H − Δ G T m (15)

A = β . E α . ⁡ exp E a R T m R . T m 2 (16)

In this context, K _*B* represents the Boltzmann constant (1.3806 × 10⁻²³ J K⁻¹), h corresponds to the Planck constant (6.6261 × 10⁻³⁴ J s^-1), T _*m* is the peak temperature in Kelvin obtained from the DTG curve, and R denotes the universal gas constant (8.314 J mol^-1 K⁻¹).

2.9 Thermodynamic parameters–effective thermodynamic approach

According to (Zhang et al., 2023) a thermodynamic model known as the Effective Thermodynamic Approach (ETA) was proposed, aimed at overcoming the limitations of conventional methods. As described in Equations 17–20, this model employs temperatures dependent on the degree of conversion (T _*α*), allowing for the variation of kinetic parameters throughout the reaction, a characteristic of complex multistage processes.

Δ H α = E a − R T α (17)

k α _ = β × E α × exp E a R T α R T α 2 (18)

Δ G α = E a + R T α ⁡ ln k B T α h k α _ (19)

Δ S α = Δ H α − Δ G α T α = R ln h k α _ k B T α − 1 (20)

In this approach, T _*α* corresponds to the temperature associated with the degree of conversion, expressed in Kelvin (K), while k α _ represents the pre-exponential factor, determined in accordance with ICTAC recommendations for conversion-dependent thermodynamic analyses. The thermodynamic parameters ΔH_α, ΔG_α, and ΔS_α refer to the changes in enthalpy, Gibbs free energy, and entropy, respectively, expressed in Kj mol^-1, kJ mol^-1, and J mol^-1 K⁻¹, and calculated based on the ETA.

2.10 Reaction mechanism

In this study, the reaction model was defined based on the expressions for f(α) and g(α) using the master plot method (Criado, 1978). This approach allows the determination of the reaction mechanism in solid-state reaction processes with considerable precision by comparing theoretical and experimental curves of Z α Z 0 , 5 as a function of conversion (α), as shown in Equation 21.

Z α Z 0.5 = f α g α f 0.5 g 0.5 = T α T 0.5 2 d α / d t d α / d t 0.5 (21)

Where f α g α f 0.5 g 0.5 represents theoretical curves and T α T 0.5 2 d α / d t d α / d t 0.5 is the experimentally obtained curve, where α = 0.5 at a 50% conversion.

2.11 Lorentzian deconvolution of DTG curves

The derivative thermogravimetric (DTG) curves obtained at different heating rates were subjected to mathematical deconvolution to separate the overlapping thermal events associated with the decomposition of moisture/light volatiles, hemicellulose, cellulose, and lignin. Initially, the experimental TG curves were numerically differentiated with respect to temperature, yielding the corresponding DTG curves. Subsequently, baseline correction was performed in OriginPro® using manually selected anchor points located before and after the degradation region, followed by software-assisted baseline subtraction. The corrected DTG curves were checked to ensure that the integrated DTG area was consistent with the total mass loss observed in the corresponding TG curve.

The DTG curves were deconvolved using nonlinear least-squares fitting with Lorentzian functions, via the Levenberg–Marquardt algorithm implemented in OriginPro® software. Each DTG curve was fitted as the sum of individual Lorentzian peaks, as described by Equation 22:where y represents the DTG signal intensity, y _*0* is the baseline shift, A corresponds to the integrated peak area, x _*c* is the center temperature of the thermal event, and ω represents the full width at half maximum.

y = y 0 + 2 A π × ω 4 x − x c 2 + ω 2 (22)

The number of Lorentzian peaks was selected according to three criteria: (i) the minimum number of peaks required to reproduce the experimental DTG profile and visible shoulders; (ii) consistency with the known degradation ranges of moisture/light volatiles, hemicellulose, cellulose, and lignin in lignocellulosic biomass; and (iii) statistical quality of the fitting, assessed by R ², reduced chi-square, and residual distribution. Four peaks were used when they provided a physically meaningful fit without systematic residual trends: a low-temperature peak associated with moisture and light volatile compounds; an intermediate peak associated mainly with hemicellulose and the initial degradation of cellulose; a main peak attributed predominantly to cellulose decomposition; and a high-temperature peak associated with gradual lignin degradation. Additional peaks were not included when they did not produce a physically meaningful improvement in the residuals or when they resulted in unrealistic peak position, width, or area (A and nca-Couce, 2022; Chang, 2020).

The initial fitting parameters, including x _*c*, ω, and A, were estimated from the local maxima and shoulders of the experimental DTG curve and subsequently refined by the iterative algorithm until convergence. The quality of the fits was evaluated using the coefficient of determination (R ^*2*), reduced chi-square (χ ^*2* red), and residual analysis, ensuring the absence of systematic trends. The absolute areas of the Lorentzian peaks were used as a measure of the relative contribution of each thermal event to the total mass loss and were normalized according to Equation 23:where P _*i* represents the percentage contribution of peak i and n is the total number of peaks considered in the fit. Due to the thermal overlap between hemicellulose and cellulose decomposition, particularly in the range of 280 °C–330 °C, the intermediate peak was interpreted as a mixed contribution rather than as an isolated component.

P i = ∣ A i ∣ ∑ i = 1 n ∣ A i ∣ × 100 (23)

3 Results and discussion

3.1 Thermogravimetric analysis

The thermal degradation of the lignocellulosic components of mallow stem residue, under an inert nitrogen (N₂) atmosphere, occurs in partially overlapping stages characteristic of pyrolytic processes. DTG curves indicated that the main mass loss occurred between 170 °C and 450 °C, associated with the simultaneous decomposition of hemicellulose and cellulose, while lignin exhibited a broader degradation profile between 450 °C and 700 °C. This behavior reflects the lower thermal stability of hemicellulose, due to its amorphous structure, and the higher thermal resistance of cellulose, related to its crystallinity, whereas lignin decomposes gradually due to its complex aromatic structure (Abioye et al., 2025; Dhyani and Bhaskar, 2018). The overlap of these stages complicates kinetic modeling using conventional single-step methods, since different reaction mechanisms occur simultaneously, resulting in variations in the apparent activation energy throughout the conversion. This multicomponent and multimechanistic nature reinforces the need to apply isoconversional methods for a more accurate description of the process.

The thermogravimetric curves and their respective derivatives are shown in Figures 1a,b. The analysis revealed four distinct stages of thermal degradation of the mallow stem under a N₂ atmosphere. Shifts in the initial (T _*i*), final (T _*f*), and peak (T _*m*) temperatures were observed as a function of the heating rate (β). In particular, the temperature corresponding to the main peak of mass loss, associated with the simultaneous degradation of hemicellulose and cellulose, increased progressively with increasing β, ranging from 187.67 °C–358.66 °C (2.5 °C min^-1), 204.51 °C–368.69 °C (5 °C min^-1), 216.53 °C–399.79 °C (10 °C min^-1), 237.80 °C–421.22 °C (15 °C min^-1), and 239.06 °C–425.76 °C (20 °C min^-1). This behavior is related to the thermal lag effect, since at lower heating rates, the system has sufficient time to complete each degradation stage, whereas at higher rates, heat transfer and reaction progression become limited (Santos et al., 2023). A similar phenomenon has been reported for other lignocellulosic biomasses, indicating that thermal displacement is characteristic of heating-rate-dependent pyrolysis processes. These results confirm the significant influence of β on degradation and reinforce the need to consider this variable in the kinetic modeling of the process.

At a heating rate of 2.5 °C·min^-1, the first thermal event occurred from room temperature up to approximately 150 °C, resulting in a mass loss of 4.96%, attributed to the release of moisture and volatile compounds characteristic of mallow biomass (Priyadharshini et al., 2023; Shah and ud din Hakim, 2025). This behavior is consistent with that observed for lignocellulosic fibers, in which the initial degradation stage occurs below 150 °C, with mass losses typically ranging between 6% and 10%, associated with the evaporation of free and adsorbed water.

The second thermal event corresponded to the most intense peak between 170 °C and 450 °C, with a mass loss of 72.55%, attributed to the overlapping decomposition of hemicellulose and cellulose. Although these components exhibit distinct thermal stability, the overlap observed in the TG/DTG curves limits the direct separation of their individual contributions at this stage, rendering interpretation by single-step models inadequate when there is variation in the apparent activation energy throughout the conversion. Under pyrolysis conditions in an inert nitrogen (N₂) atmosphere, hemicellulose is typically associated with the formation of low-molecular-weight oxygenated compounds, as well as the release of H₂O, CO₂, and CO, while cellulose tends to generate anhydrosugars, notably levoglucosan, in addition to CO, CO₂, and H₂O, concurrently with the formation of the carbonaceous fraction (char). Lignin, on the other hand, exhibits degradation distributed over a broad temperature range and is recognized as the main precursor of phenolic compounds during pyrolysis (Kumagai et al., 2021; Lu and Gu, 2022).

The third and fourth thermal events occurred between 358 °C and 700 °C, resulting in a total mass loss of 21.26%, and can be distinguished based on the DTG profile. In the range of 358 °C–450 °C, a secondary shoulder peak is observed, attributed to the final stage of the decomposition of the remaining cellulose and the onset of lignin degradation, while in the range of 450 °C–700 °C, the gradual decomposition of lignin and the rearrangement of the carbonaceous residue (char) predominate, with the formation of aromatic and phenolic compounds and light gases under an inert nitrogen (N₂) atmosphere. The main degradation stage, however, occurred between 187 °C–358 °C, with a mass loss of 72.55%, predominantly associated with the overlapping decomposition of hemicellulose and cellulose, the latter being structurally more stable and, therefore, generally associated with higher activation energy values. Considering that this region accounts for the largest fraction of the conversion, the kinetic analysis was conducted by taking this dominant event as representative of the overall process, an approach consistent with biomass studies in which the main degradation stage is used as a reference when secondary events have a smaller relative contribution.

The TG/DTG-derived analysis indicated a moisture content of 2.83%, a volatile matter content of 72.56%, a fixed carbon content of 5.43%, and a final residue of 5.570% at 700 °C under N₂ atmosphere. This final residue corresponded to approximately 0.274 mg from an initial sample mass of 4.919 mg. Since the analysis was performed under an inert atmosphere, this value should be interpreted as a final carbonaceous/mineral residue rather than as true ash content. The low moisture content observed is typical of pre-dried samples and is favorable for thermochemical applications, as it reduces the energy required for water evaporation during the thermal process. The high volatile matter content highlights the predominantly lignocellulosic nature of the biomass, indicating high thermal reactivity and greater release of gaseous and condensable products during pyrolysis. In turn, the final carbonaceous/mineral residue may influence char formation and secondary reactions during thermal decomposition.

Based on the corrected TG/DTG-derived parameters, the higher heating value was recalculated using the empirical HHV correlation, resulting in an estimated value of 13.19 MJ kg^-1. This value should be interpreted as an estimated energetic indicator rather than as a direct calorimetric measurement, since the final residue obtained under N₂ may contain both carbonaceous char and inorganic mineral matter.

3.2 Gas evolution by TG-MS

Figure 2 shows the ion current profiles as a function of temperature for selected m/z ratios during the pyrolysis of mallow stem residue. These profiles provide qualitative information on the evolution of volatile fragments released during thermal degradation. However, since TG-MS monitoring was based on nominal m/z signals, the assignments presented in this section should be interpreted as tentative fragment attributions rather than definitive molecular identification.

The TG-MS profiles indicate that volatile release occurs mainly between 200 °C and 400 °C, with maximum intensities concentrated between approximately 280 °C and 340 °C, in agreement with the main mass-loss stage observed in the TG/DTG curves. The most intense and reliable signals were observed at m/z = 18 and m/z = 44, which were assigned to H₂O and CO₂, respectively. The m/z = 18 profile shows an initial contribution at low temperature, associated with moisture release, followed by a more intense peak in the main degradation region, indicating dehydration reactions during the decomposition of polysaccharide-rich fractions. The m/z = 44 signal shows a broad evolution profile, consistent with CO₂ formation from decarboxylation and deoxygenation reactions during biomass pyrolysis (Xue et al., 2025).

Hemicellulose, mainly composed of xylans, tends to release a higher fraction of light oxygenated volatile compounds and gases such as H₂O, CO₂, and CO (Yang et al., 2007), including species derived from acetyl groups, such as acetic acid (m/z = 60), as well as furan compounds (Saires et al., 2024). Cellulose, the predominant component of the lignocellulosic fraction, contributes significantly to the main degradation peak and forms anhydrosugars, as well as characteristic carbohydrate fragments with ions around m/z 73–74, simultaneously with the formation of carbonaceous residues (char) (Chang et al., 2026; Kline et al., 2020). In addition, the signals at m/z = 15, associated with CH₄/CH₃ ⁺, and m/z = 30, attributed to formaldehyde or light oxygenated fragments, reached maxima near the main degradation peak, indicating intense volatile formation during primary pyrolysis (Barros et al., 2019).

The profiles observed for m/z = 53, 55, 57, 60, 64, 74, and 81 were interpreted more cautiously due to possible overlapping fragments. These signals are less intense than m/z = 18 and m/z = 44 and are mainly concentrated in the 250 °C–350 °C region, suggesting the release of organic fragments during the active pyrolysis stage. The ions at m/z = 53, 55, and 57 may be related to unsaturated hydrocarbon fragments, furan-derived fragments, or other organic species formed from carbohydrate degradation. The signal at m/z = 60 may be associated with acetic acid-related fragments or other oxygenated species derived from hemicellulose. The m/z = 74 signal may correspond to oxygenated organic fragments, including possible C₃H₆O₂ isomers. However, the assignments of m/z = 46, 64, and 81 are particularly uncertain because these nominal masses may arise from more than one molecular source or from overlapping fragmentation pathways (Chang et al., 2026).

Supplementary Table S3, available in the Supplementary Material, lists the main m/z ratios identified in the TG-MS profiles of mallow stems. This classification allows us to correlate the evolution of gaseous species with the different stages of thermal decomposition.

Studies using TG-MS on biomass and waste materials also report broad H₂O and CO₂ evolution profiles, which are generally interpreted as the combined result of moisture release, dehydration, deoxygenation, decarboxylation, and decarbonylation reactions throughout the degradation process (Xue et al., 2025). Thus, the TG-MS results support the multistage thermal degradation behavior observed by TG/DTG, while the specific assignment of minor fragments remains tentative due to the absence of complementary FTIR or GC-MS confirmation.

3.3 Kinetic parameters

In this study, the FWO, KAS, and Starink isoconversional methods, as well as the nonlinear isoconversional Vyazovkin method, were applied. The apparent activation energy values obtained using these methods at different conversion levels are presented in Table 1. These methods are suitable for multistage systems such as lignocellulosic biomass, in which pyrolysis involves overlapping reactions of hemicellulose, cellulose, and lignin, as evidenced by TG/DTG and TG–MS analyses. The activation energy for each conversion level (α) was obtained from the slopes of the linear plots, which exhibited consistent parallelism and coefficients of determination of R² ≈ 0.99.

TABLE 1. Kinetic parameters of mallow stem residue obtained using isoconversional and nonlinear isoconversional methods.

Bold values represent the average apparent activation energy (E_a) obtained for each isoconversional method. The uncertainties reported for the average E_a values of FWO, KAS, and Starink were estimated from the regression statistics of the linear isoconversional plots. For the Vyazovkin method, the average E_a value is reported without regression-based uncertainty because this method is based on nonlinear numerical minimization rather than linear fitting.

This parallelism indicates the absence of abrupt changes in the degradation mechanism within the analyzed temperature range. The main stage of thermal degradation, identified in the DTG curves between approximately 280 °C–370 °C, corresponds to the conversion range of α ≈ 0.20–0.75, as observed in the conversion curves shown in Figure 3 This range is predominantly associated with the decomposition of hemicellulose and cellulose.

The high linearity of the ln (β) versus 10⁻³ T^-1 plots indicates mathematical consistency in the application of isoconversional models within the analyzed conversion range, which does not necessarily imply significant reaction progress at very low temperatures. The dominant kinetic contribution is concentrated in the range corresponding to the degradation of hemicellulose and cellulose, whose polysaccharide structures differ from those of aromatic lignin, reflecting distinct reaction mechanisms and activation energy profiles throughout the conversion.

As shown in Figure 4, the nearly parallel lines indicate a moderate variation in the apparent activation energy within the dominant conversion range, suggesting that hemicellulose and cellulose are the primary contributors. In the present study, an increase in the apparent activation energy (E _*a*) was observed up to α ≈ 0.5, a behavior similar to that reported for wheat straw, with values ranging from 125 to 193 kJ mol^-1 at α = 0.1–0.8 (Rathore et al., 2023; Vyazovkin et al., 2011) where the increase up to α ≈ 0.5 is associated with the progressive decomposition of polysaccharides. At initial conversions in α = 0.1–0.45, slight deviations are observed, which may reflect the contribution of non-reactive processes such as drying and removal of external and physically bound moisture, as well as extractives, before the predominance of active pyrolysis; therefore, the parameters estimated in this range should be interpreted as apparent activation energy values, and the main kinetic discussion should focus on the α range associated with the dominant devolatilization stage (Hossain et al., 2025; Mishra and Mohanty, 2020). This behavior stems from the structural differences among the components of biomass. Hemicellulose, with its amorphous and branched structure, lower degree of polymerization, and presence of acetylated groups, decomposes more readily, beginning to lose mass at lower temperatures. Cellulose, due to its semicrystalline structure and strong network of hydrogen bonds, undergoes a more concentrated and energetically demanding decomposition. Lignin, on the other hand, formed by three-dimensionally intertwined aromatic units, degrades more slowly and in a distributed manner, contributing to thermal overlap and the absence of abrupt kinetic transitions (Santos et al., 2023; Dhyani and Bhaskar, 2018; Wang et al., 2022).

The FWO, KAS, and Starink isoconversional models were originally developed to estimate the activation energy without assuming a specific reaction mechanism and are widely applied to both single-step reactions and complex systems. In multistage processes, such as the pyrolysis of lignocellulosic biomass, these methods provide an apparent activation energy that depends on the degree of conversion, reflecting the combined contribution of the different constituents. Although they do not allow the discrimination of individual mechanisms, they are considered robust for describing heterogeneous systems when applied within conversion ranges associated with the dominant reaction step.

The average activation energy was 126.39 kJ mol^-1 (FWO), 122.95 kJ mol^-1 (KAS), and 123.43 kJ mol^-1 (Starink). These values should be interpreted as apparent activation energy values for the conversion range associated with the dominant stage of pyrolysis, in which there is an overlap of contributions from hemicellulose and cellulose. In comparative terms, the literature typically reports lower E _*a* values for hemicellulose and higher values for cellulose, while lignin exhibits E _*a* values widely distributed throughout the conversion; thus, the values obtained (∼123–126 kJ mol^-1) are consistent with a multicomponent process dominated by polysaccharides in the analyzed range, rather than the isolated decomposition of a single constituent.

The activation energy (E _*a*), pre-exponential factor (A), and correlation coefficient (R ^*2*) across the biomass conversion range are presented in Table 2. The average activation energy values obtained by the FWO, KAS, and Starink methods were 126.39 ± 0.59, 122.95 ± 0.65, and 123.43 ± 0.63 kJ mol^-1, respectively. These uncertainties were estimated from the regression statistics of the linear isoconversional plots. For the nonlinear Vyazovkin method, the average Ea was 123.38 kJ mol^-1, showing close agreement with the KAS and Starink methods.

TABLE 2. Comparison of thermal degradation behavior and activation energy values of mallow stem residue with selected lignocellulosic biomasses reported in the literature.

In general, the activation energy for the mallow stem residue decreases gradually as the conversion α increases, ranging from approximately 133 kJ mol^-1 at α = 0.10 to approximately 104 kJ mol^-1 at higher conversions α = 0.90. This reduction reflects the complexity of the pyrolysis process, in which the initial decomposition of hemicellulose and cellulose requires higher energy than the later stages, which may exhibit a decreasing trend in Ea at higher conversions due to the multistage nature of the degradation and the increasing contribution of secondary reactions and more recalcitrant fractions (Fang et al., 2026; Zsinka et al., 2024).

According to the ICTAC (International Confederation of Thermal Analysis and Calorimetry), the accuracy of isoconversional methods follows the order Starink > KAS > FWO. In the study reported in (Rathore et al., 2023), the authors showed that the variation between the average activation energies obtained by FWO, KAS, and Starink was ≤1 kJ mol^-1, which underscores ICTAC’s recommendation for the comparative use of these methods. Furthermore, it is noted that KAS tends to provide more accurate values than FWO due to the integral approximation used in the FWO method (Doyle approximation), while the Starink method combines different formulations and reduces this bias.

In the same study, at α = 0.10, the values are very close across the models (FWO: 125.21; KAS: 127.52; Starink: 125.25 kJ mol^-1), illustrating this convergence. Furthermore, STK and KAS presented very similar results, consistent with the ICTAC description, reinforcing that the results obtained by the STK method can be considered more reliable for describing the pyrolysis kinetics of mallow stem residue. In the study of polycultivated microalgal biomass, KAS and FWO provide very similar average E _*a* values, while the STK fits show less variation across the conversion range (Mohit and Remya, 2024).

Vyazovkin’s nonlinear isoconversional method, although it does not require constant activation energy and employs a more flexible modeling approach, yielded an average E _*a* of 123.38 kJ mol^-1, which is virtually identical to the value obtained by the Starink method. This result demonstrates that, for this biomass and under these experimental conditions, the differences between methods are small. This evidence is consistent with recent studies indicating that, in multistage processes, the variation of E _*α* with α is expected and that isoconversional methods are appropriate for capturing this dependence without imposing a single mechanism; on the other hand, model-fitting approaches can introduce biases when applied assuming a single reaction stage (Açıkalın et al., 2024; Vyazovkin et al., 2020).

Table 2 compares the activation energy values obtained for mallow stem residue with those reported for other lignocellulosic biomasses. The average Ea values found in this study, ranging from 122.95 to 126.39 kJ mol^-1, were higher than those reported for corn stover, rapeseed straw, cotton straw, and flax straw (Kim et al., 2022; Fernandez et al., 2024). This suggests that mallow stem residue requires a higher apparent energy input for thermal degradation than several straw-type residues, possibly due to its fibrous structure and the contribution of cellulose- and lignin-rich fractions during the main devolatilization stage.

On the other hand, the E _*a* values of mallow stem residue were lower than the broader ranges reported for bamboo powder, pineapple residue, and banana pseudo-stem (Casal et al., 2023; Özsin, 2025). Therefore, mallow stem residue presents an intermediate kinetic behavior among the compared lignocellulosic feedstocks. Although the devolatilization regions reported in the literature may vary according to the criteria adopted in each study, the comparison indicates that mallow stem residue falls within the typical range of biomass pyrolysis and supports its potential for energy-related applications.

The variation of activation energy as a function of the degree of conversion for the pyrolysis of mallow stem waste is shown in Supplementary Figure S1 of the Supplementary Material for the different isoconversion methods (FWO, KAS, STK, and VYZ). All models exhibit a similar trend, with values close to 130 kJ mol^-1 in the initial stages (α < 0.4) and a progressive decrease to approximately 105 kJ mol^-1 at α ≈ 0.9. This behavior is associated with structural differences between the biomass constituents. At initial conversions, the cleavage of glycosidic bonds in hemicellulose and the less crystalline regions of cellulose require more energy due to the breaking of C–O–C bonds and the depolymerization of polysaccharide chains. As the conversion progresses, reactions involving structural reorganization, aromatization, and condensation of the remaining carbonaceous material predominate; these processes involve energetically distinct mechanisms and may exhibit lower apparent activation energy, resulting in a progressive decrease in E _*a* (Lei et al., 2023).

In contrast, pineapple waste and banana pseudostems exhibited higher Ea ranges of 159–335 kJ mol^-1 and 169–364 kJ mol^-1, respectively, for FWO (Wang et al., 2022). These higher values are associated with a greater fraction of structurally crystalline cellulose and a higher lignin content, whose three-dimensional aromatic units confer greater thermal stability. The semicrystalline nature of cellulose, compared to the amorphous and less organized structure of hemicellulose, requires more energy to break intermolecular bonds, explaining the observed differences in activation energy.

3.4 Thermodynamic parameters

The thermodynamic parameters discussed in this section were derived from TG-based kinetic data and should therefore be interpreted as apparent thermodynamic descriptors of the pyrolysis process. In solid-state biomass decomposition, pyrolysis is driven by continuous external heating, and the calculated ΔH, ΔG, and ΔS values are associated with the formation of the activated complex under the assumptions of the kinetic model used. Thus, these parameters should not be interpreted as absolute thermodynamic state functions or as direct equilibrium criteria.

3.4.1 Conventional approach

Conventional thermodynamic parameters were calculated at a heating rate of 10 °C min^-1 using the FWO, KAS, STK, and VYZ isoconversion models. The complete conversion-dependent values of ΔH, ΔG, and ΔS are presented in Supplementary Table S1, while the main trends are discussed in this section in Table 3.

TABLE 3. Comparison of lignocellulosic component contents of mallow stem residue obtained by chemical analysis and DTG deconvolution at 10 °C min^-1.

The average enthalpy values were ΔH ≈ 120–121 kJ mol^-1, which are close to the corresponding activation energy values, indicating that the RTm term had a limited influence on the calculation. Comparable values have been reported for bamboo pyrolysis, with ΔH ranging approximately between 133 and 142 kJ mol^-1 in the main degradation stage, demonstrating similar thermodynamic behavior among lignocellulosic biomasses (Lei et al., 2024). The Gibbs free energy values showed an average of approximately 174 kJ mol^-1, indicating that the formation of the activated complex is energetically demanding under the experimental pyrolysis conditions. However, in thermally activated solid-state decomposition, positive ΔG should not be interpreted only as a classical equilibrium-based spontaneity criterion. Instead, it reflects the need for continuous thermal activation to promote bond cleavage, volatile release, and structural rearrangement in the biomass matrix. Similar ΔG ranges have been reported for Camellia oleifera residues and biomass/plastic co-pyrolysis systems (Mariyam et al., 2023; Xia et al., 2024), supporting the consistency of the values obtained for mallow stem residue. The entropy values (ΔS) are consistently negative, ranging from ≈ − 90 to −95 J mol^-1·K⁻¹, suggesting a more ordered activated state compared to the initial state, associated with structural rearrangements during the process, as observed in the case of bamboo, where negative ΔS accompanies the main stages of thermal degradation (Lei et al., 2024).

When comparing these models, the FWO method tends to yield slightly higher values for ΔH and E _*a*, while KAS produces lower averages; Starink and Vyazovkin provide estimates that are very close to one another, consistent with accuracy trends reported in recent studies and supported by international recommendations (Ben Abdallah et al., 2023; Lei et al., 2025). In practical terms, the positive ΔH values indicate that heat input is required for the main decomposition reactions, while positive ΔG values reflect the energetic demand associated with activated complex formation. The negative ΔS values suggest a more constrained activated state compared with the initial solid structure, which may be related to structural rearrangements and reduced molecular mobility during solid-state degradation (A and nca-Couce, 2022).

3.4.2 Effective thermodynamic approach

The effective thermodynamic parameters estimated using the FWO, KAS, STK, and VYZ methods also showed consistent behavior across the conversion range. The complete conversion-dependent values are provided in Supplementary Table S2. In the main text, the discussion focuses on the average values and trends observed during pyrolysis.

The effective thermodynamic parameters estimated using the FWO, KAS, STK, and VYZ isoconversion methods show high consistency across the conversion range. For mallow stems, the average values obtained for ΔH ≈ 118–121 kJ mol^-1, ΔG ≈ 154–155 kJ mol^-1, and ΔS ≈ −55 to −61 J mol^-1, indicating an endothermic process and the formation of a more varied activated complex (Lei et al., 2024), when studying the pyrolysis of bamboo, observed ΔH between 115 and 125 kJ mol^-1, ΔG between 150 and 165 kJ mol^-1, and a negative ΔS between −40 and −80 J mol^-1 K⁻¹ throughout the conversion. Similar results were also reported by (Mariyam et al., 2023) for C. oleifera residues, in which ΔG varied approximately between 170 and 180 kJ mol^-1 during pyrolysis. Thus, the parameters obtained for the mallow stem fall within the typical range observed for lignocellulosic biomass, confirming the thermodynamic consistency of the process.

This behavior is typical of the thermal decomposition of lignocellulosic biomass, indicating that the process occurs through thermally activated global mechanisms that require continuous heat input. That depend on a continuous supply of thermal energy and are characterized by multiple overlapping reaction steps (Ivanovski et al., 2021; Patidar et al., 2022; Silva et al., 2023).

The ΔH values showed a decreasing trend as the conversion increased, regardless of the method used. For the FWO method, ΔH ranged from 128.70 kJ mol^-1 (α = 0.10) to 103.70 kJ mol^-1 (α = 0.90), corresponding to an approximate 19% decrease over the course of the conversion. The KAS, STK, and VYZ methods yielded very similar average values, around 118 kJ mol^-1, with differences of less than 3% among them. These results are consistent with the traditional method, in which ΔH is estimated from E _*a* − RT, yielding values close to the activation energy obtained by the kinetic models.

The progressive decrease in enthalpy with increasing conversion observed in this study is consistent with results reported in the literature for lignocellulosic biomass. ΔH decreases from 128.70 kJ mol^-1 (α = 0.10) to 103.70 kJ mol^-1 (α = 0.90), indicating that the initial stages of pyrolysis require higher energy input than the later stages. A similar trend was observed in the pyrolysis of wheat straw, where ΔH values ranged from approximately 135–142 kJ mol^-1 in the conversion range of α = 0.20–0.60, reflecting changes in decomposition mechanisms throughout the thermal process (Lei et al., 2024).

This behavior is typical of the pyrolysis of lignocellulosic biomass, as decomposition involves sequential contributions from hemicellulose, cellulose, and lignin, whose chemical structures exhibit different thermal stabilities and bond energies, resulting in variations in kinetic and thermodynamic parameters throughout the conversion. At high conversion levels, the decrease in enthalpy is associated with the progressive degradation of residual lignin and secondary reactions involving less ordered carbonaceous structures. Studies report typical ΔH values on the order of ∼164 kJ mol^-1 in the main stages, with a reduction in the energy barrier in the final stages of char degradation, which can reach significantly lower values, reflecting the lower energy requirement for the decomposition of remaining aromatic structures (Patil and Ku, 2024; Patil et al., 2023; Shen et al., 2025).

Comparable results have also been reported by authors who investigated the pyrolysis of lignocellulosic fibers. In the studies by (Li et al., 2023), when analyzing bamboo residues and their structural constituents, they observed effective ΔH values in the same order of magnitude as those obtained in this study, with a progressive decrease associated with the progress of the conversion. Similarly (Yiga et al., 2025), found that the decrease in ΔH during the thermal decomposition of rice husk fibers reflects the transition between more organized fractions and less complex residual structures, demonstrating thermodynamic behavior typical of lignocellulosic fibrous materials.

Gibbs free energy has also been reported to be positive throughout the conversion process in various studies on biomass pyrolysis, indicating the non-spontaneous nature of thermal decomposition and the need for an external supply of thermal energy. For example, in (Vafakish et al., 2023), the ΔG value calculated for the pyrolysis of flax straw was approximately 162.6 kJ mol^-1, consistent with the ranges observed in thermodynamic analyses of biomass. Similarly (Bongomin et al., 2024), observed positive ΔG values for various lignocellulosic wastes, highlighting that such processes require continuous energy input to proceed. Furthermore, studies on tropical biomass indicate that positive ΔG reflects the predominant endergonic energy demand during the pyrolysis of lignocellulosic materia (Yu et al., 2025).

The gradual increase in ΔG observed in this study as the conversion progresses indicates that the formation of the activated complex becomes progressively less favorable in the final stages of pyrolysis. Although the activation energy shows a slight decrease throughout the conversion, the increase in ΔG is mainly associated with the increasingly negative values of ΔS, which increases the term −TΔS in the thermodynamic relationship ΔG = ΔH − TΔS. This behavior is typical of the decomposition of lignocellulosic biomass, in which the final stages involve structural reorganization and the formation of more ordered char. This behavior was also reported by (Ivanovski et al., 2021), who attributed the increase in ΔG at high conversions to the greater thermal stability of the residual fractions of the biomass.

The activation entropy values were negative across the entire conversion range, ranging approximately from −22 to −105 J mol^-1·K⁻¹, with average values between −55 and −61 J mol^-1·K⁻¹. The negative values of ΔS indicate that the transition state associated with thermal decomposition is more ordered than the reactants, suggesting a mechanism controlled by structural constraints and the solid nature of the system.

Thus, although in the present study the conventional method yielded higher ΔG values and more negative ΔS values than the effective thermodynamic method, contrary to what was observed by (Salami et al., 2025) for sotol bagasse both studies agree on a key point: the differences between the methods are concentrated primarily in ΔG and ΔS, while ΔH remains very similar. In this study, the mean values of ΔH ranged only from 117.96 to 121.40 kJ mol^-1, whereas ΔG decreased from 173.96–175.53 kJ mol^-1 in the conventional method to 154.90–155.05 kJ mol^-1 in the effective method, and ΔS became less negative, changing from −89.07 to −94.98 J mol^-1 K⁻¹ to −54.94 to −60.86 J mol^-1 K⁻¹. In contrast, the effective thermodynamic method yields more moderate values for these parameters, suggesting a more realistic representation of the energy evolution of the thermal decomposition process and the progressive complexity of the reaction mechanism in lignocellulosic systems.

3.5 Reaction mechanism

Pyrolysis of mallow stem residue reveals the presence of multiple reaction steps, making it impractical to attribute the entire process to a single mechanism. As shown in Figure 5, using Criado’s master plot method, the experimental curve for α = 0.1–0.9 at a heating rate of 10 °C min^-1 was compared with theoretical models such as diffusion (Dn), random nucleation (Rn), and phase contraction (Fn). In the initial stages of conversion, between α ≈ 0.1 and 0.3, corresponding to the temperature range of 200 °C–280 °C, a better fit to the three-dimensional diffusion model (D4) is observed, indicating partial control by diffusion of the products through the solid matrix. In agreement with the TG–MS results presented earlier, this region is characterized by the predominance of water release and the onset of carbon dioxide formation, associated with the initial decomposition of hemicellulose. The low intensity of heavier fragments reinforces the limited formation of complex volatiles, indicating a regime dominated by light species in a structure that is still relatively intact. This behavior is consistent with observations reported for cellulose-rich biomasses, such as sugarcane bagasse, in which diffusion controls the initial devolatilization (Libo et al., 2023).

For α between approximately 0.3 and 0.6, the experimental curve showed good agreement with R2/R3 models, characteristic of reactions controlled by the solid-product interface, in which decomposition proceeds from the surface toward the interior of the particle, a process typical of cellulose degradation. In contrast (Fischer et al., 2024), reported a better fit for nth-order models (F1–F5) in different biomasses, indicating predominant control by the global chemical reaction. This difference suggests that, in mallow stem residue, pyrolysis is more influenced by the structural evolution of the lignocellulosic matrix and by char formation.

At α > 0.6, the experimental behavior deviates from diffusion models and approaches mechanisms controlled by surface reactions or boundary-layer processes, indicating that the decomposition rate is now governed by the breakdown of more stable aromatic structures in lignin. Unlike hemicellulose and cellulose, which have more accessible chains and degrade rapidly, lignin has a highly cross-linked three-dimensional structure with more resistant C–C and C–O bonds, making its degradation slower and energetically more complex. This pattern has also been observed in studies of the torrefaction and pyrolysis of industrial lignin, in which the final stages of the thermal process are dominated by surface reactions associated with the formation and reorganization of char (Menezes et al., 2024).

These results indicate that the pyrolysis of mallow stem residue involves a transition from a diffusion mechanism to nucleation and, subsequently, surface-controlled decomposition as the reaction progresses. In agreement with the TG–MS data presented, it is observed that, in the initial stages, light gases such as water and carbon dioxide predominate, which diffuse more easily through the solid matrix, justifying the diffusive regime. As the conversion progresses, there is an increase in the formation of organic compounds with higher molecular weights, whose release is more restricted by the solid structure, favoring the transition to nucleation and growth mechanisms. In more advanced stages, decomposition becomes controlled by reactions on the surface of the remaining carbonaceous material, associated with the slower release of these heavier volatiles. This behavior is consistent with the multistage nature of lignocellulosic biomass degradation described in the literature. Similarly (Libo et al., 2023), observed a predominance of nth-order kinetic models in sugarcane bagasse during devolatilization, and (Menezes et al., 2024) reported that, in the final stages, decomposition is controlled by the more stable aromatic structures of lignin, a behavior consistent with that observed for mallow.

3.6 Lorentzian deconvolution of DTG curves

The deconvolution of DTG curves was adopted as an alternative approach in this study for estimating the relative composition of hemicellulose, cellulose, and lignin. In DTG curves of lignocellulosic biomass, degradation events often appear superimposed and can therefore be described by the sum of peak functions, allowing the integrated areas to be related to the pseudocomponent fractions of the material (Kim et al., 2022). Recent studies explicitly apply Lorentzian functions to separate overlapping reaction steps from dα/dT curves, obtaining individual peaks via nonlinear fitting using the Levenberg–Marquardt algorithm, which justifies the use of the same rationale in this study (Fernandez et al., 2024).

Supplementary Figure S2 of the Supplementary Material shows the deconvolution results for a heating rate of 10 °C min^-1, as this condition provides good visual resolution of the events and a satisfactory model fit. The peak separation method using symmetric functions, including Lorentzian functions, continues to be widely used in recent thermochemical analyses, either directly (Lorentzian) or in hybrid formulations (Gauss-Lorentz/Voigt), particularly when the objective is to quantify the contributions of overlapping thermal events to the derived signal (Casal et al., 2023; Özsin, 2025).

The cumulative fit adequately tracks the experimental DTG curve across the entire temperature range, with R ² = 0.98665 at a heating rate of 10 °C·min^-1, indicating that the decomposition can be described as the sum of four major events. The first peak, centered at approximately 63 °C, is consistent with the loss of physically adsorbed moisture and light volatiles.

Next, the peak at approximately 300 °C represents a broad event associated primarily with hemicellulose, but likely involving a contribution from the initial fraction of amorphous cellulose, as overlap in this region is common in lignocellulosic systems. The peak of highest intensity, centered at approximately 346 °C, corresponds to the main degradation of cellulose, while the peak at approximately 484 °C, which is broader and occurs at higher temperatures, is related to the gradual decomposition of lignin and more resistant fractions, whose degradation extends to higher temperatures. Recent studies applying Lorentzian deconvolution models to lignocellulosic residues also report similar behavior, with a moisture and light volatiles event at low temperatures and main events associated with biopolymers in progressively higher temperature ranges, with lignin contributing at higher temperatures and exhibiting a broader signal (Sahu and Gangil, 2025).

Since the DTG signal is negative (mass loss per unit temperature), the calibration yields negative A values, and ∣A∣ is used for percentage quantification. The peak width reinforces the physicochemical interpretation: the peak at ∼300 °C has greater amplitude and width, which is consistent with a composite event superimposed on hemicellulose and the initial cellulose composition at ∼346 °C, where it is narrower and more intense, suggesting a well-defined dominant event.

The differences between the contents obtained by chemical analysis and thermal deconvolution stem from the different principles underlying the methods. Chemical analysis defines the fractions through extraction, which may leave some of the insoluble lignin in the residue or redistribute components. DTG deconvolution, on the other hand, is based on thermal decomposition bands, which show significant overlap, causing part of the cellulose to decompose alongside hemicellulose and lignin to be distributed across different temperature regions. Thus, the thermal values represent apparent fractions, explaining the higher estimate of cellulose and the lower estimate of lignin observed.

Table 3 presents a comparison between the hemicellulose, cellulose, and lignin contents of the mallow stem residue obtained by chemical analysis and those estimated by deconvolution of the TG curves. Good agreement is observed for the hemicellulose fraction, whose values were virtually identical between the two methods after peak area correction. However, more pronounced differences were observed for cellulose and lignin. Part of the cellulose may be partially incorporated into the intermediate peak near approximately 300 °C, especially when these peak exhibits greater width, highlighting the well-known thermal overlap between hemicellulose degradation and the initial cellulose fraction in the DTG curves. Recent thermogravimetric studies demonstrate that symmetric functions, such as Lorentzian and Gaussian, can provide an adequate preliminary approximation for the decomposition of biomass, but they have limitations when mass loss events are strongly overlapped, generating broad peaks and shoulders that are not fully represented by pure symmetric functions (Paredes et al., 2024; Cao et al., 2024).

In such cases, authors recommend the use of asymmetric functions or complementary modeling strategies, such as the Fraser-Suzuki or Bi-Gaussian models, or even the division of the temperature range into sub-events to more strictly separate the contributions of hemicellulose and cellulose (Paredes et al., 2024; Athparia et al., 2026). These approaches, although more complex, better represent the actual variability of thermal degradation in biomass with heterogeneous composition, justifying methodological caution when interpreting a single broad intermediate peak when using symmetric deconvolution.

4 Conclusion

This study demonstrated that the stem residue of mallow (Urena lobata Linn.) is a thermally reactive lignocellulosic byproduct with potential for energy applications. The main degradation step occurred between 200 °C and 400 °C, primarily associated with the decomposition of hemicellulose-rich and cellulose fractions. The final residue corrected under a N₂ atmosphere was 5.570%, and the estimated higher heating value was 13.19 MJ kg^-1. The FWO, KAS, Starink, and Vyazovkin methods provided consistent activation energy values, with average Ea values between 123 and 126 kJ mol^-1, indicating conversion-dependent thermal degradation. Thermodynamic and reaction mechanism analyses confirmed that the process is thermally activated and occurs through multiple degradation steps. Therefore, mallow stem residue can represent an alternative biomass for energy-related applications and for the valorization of an underutilized fibrous byproduct from amazon. However, further studies involving ash composition, analysis of released gases, product characterization, and reactor-scale experiments are essential to confirm its practical applicability.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/ffuel.2026.1873774/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Nikolaos C. Kokkinos, Democritus University of Thrace, Greece

Reviewed by: Kanti Kumar Athankar, IPS Academy Institute of Engineering and Science, India

Hesti Wijayanti, Universitas Lambung Mangkurat Fakultas Teknik, Indonesia