Sustainable Thermoplastic Starch Biocomposites from Coffee Husk and Mineral Residues: Waste Upcycling and Mechanical Performance

Abstract

Thermoplastic starch (TPS) is a biodegradable polymer from renewable sources, but its limited mechanical and thermal properties restrict wider industrial use compared to petroleum-based plastics. In this study, TPS-based biocomposites were developed and optimized by incorporating agricultural and mineral Residues: coffee husks (CH), potassium feldspar (PF), and Bahia Beige marble (BB) as reinforcements. Mechanical, thermal, and morphological characterizations were carried out, and a simplex–lattice mixture design was applied to optimize the formulations. The 70/20/5/5 (TPS/CH/PF/BB, wt.%) composition achieved the highest tensile strength (2.0 MPa) and elastic modulus (70.2 MPa), while the 90/0/5/5 formulation showed superior impact resistance. FTIR and SEM analyses confirmed effective filler dispersion and strong matrix–filler interactions. Scheffé polynomial models (R² > 87%) accurately predicted performance, highlighting the reliability of the statistical approach. From a sustainability perspective, this work demonstrates that upcycling coffee husks and mineral residues into TPS-based biocomposites contributes to waste reduction, landfill diversion, and the development of cost-effective biodegradable materials. The proposed systems offer potential for eco-friendly packaging and agricultural applications, reducing dependence on fossil-based plastics and mitigating the environmental footprint of polymer industries. Statistical optimization further enhances efficiency by minimizing experimental waste. Moreover, this research supports circular economy strategies and provides scalable, sustainable solutions for waste valorization.

1. Introduction

Native starch is unsuitable for direct plastic applications due to its high-water vapor permeability and poor mechanical performance. To overcome these drawbacks, starch is typically plasticized with additives such as water or glycerol under controlled heat and shear. This process disrupts the granular structure and promotes gelatinization. It results in thermoplastic starch (TPS), where amylose dissolves and amylopectin crystallites melt. Plasticizers play a crucial role by reducing intermolecular forces, thereby increasing chain mobility, improving flexibility, and reducing brittleness. Nevertheless, TPS films remain mechanically and functionally inferior to petroleum-based plastics. This limitation reinforces the need for further modifications to render TPS viable for industrial use [1,2,3,4,5].

In this context, starch-based biocomposites emerge as a promising strategy to enhance both mechanical and barrier properties while retaining biodegradability. Their attractiveness lies in the combination of low cost, renewable origin, and environmental sustainability. Furthermore, the addition of natural fibers or mineral fillers allows tuning of specific performance attributes, making starch-based composites competitive with conventional polymeric materials in multiple applications [5,6,7].

Particularly, lignocellulosic residues such as coffee husks represent valuable reinforcements. Coffee husk, a by-product of coffee processing, contains approximately 55–57% cellulose and 22–35% lignin. It also offers a rich source of bioactive compounds and structural fractions. This composition positions it as an abundant and sustainable raw material, consistent with the growing trend of valorizing agricultural residues. Such strategies aim to reduce waste and improve the environmental footprint of polymer-based systems [8,9].

Complementing this perspective, other studies have highlighted the potential of industrial residues for composite reinforcement. As reported by Kumar et al. [10], Barros et al. [11], and Chagas et al. [12], the ornamental stone industry generates substantial amounts of both solid waste and stone slurry. Solid residues typically result from rejects in mining or processing plants, while stone slurry emerges as a semi-liquid suspension of fine particles produced during sawing and polishing. These materials are often disposed of in landfills, which contributes to environmental degradation. Addressing this challenge, Silva et al. [13] formulated PLA-based hybrid composites incorporating coffee husk, potassium feldspar, and Bahia Beige marble. FTIR, hardness, contact-angle, density measurements, SEM, XRF, and statistical analyses indicated favorable overall performance. Coffee husks notably doubled the degradation rate of PLA, while calcium- and potassium-rich fillers enhanced soil quality. These findings evidenced multifunctional environmental benefits. This work also highlighted the feasibility of fully biodegradable, thermoformed, multilayered composite capsules designed for sustainable coffee packaging.

Building on these results, the present study proposes the use of thermoplastic starch (TPS) as the matrix for Biocomposites, an approach not yet explored with fillers such as coffee husks, potassium feldspar, and Bahia Beige marble. The design of these TPS-based Biocomposites follows a framework that links filler selection and composition to target properties, including tensile strength, thermal stability, and processability. The intended applications, such as biodegradable packaging, define the required performance specifications, guiding both formulation and processing. By combining starch, a widely available biodegradable polymer, with agro-industrial residues and mineral fillers, this research introduces a novel pathway for developing sustainable and functional materials, expanding the knowledge on Biocomposite design and enhancing the industrial viability of starch-based alternatives.

2. Materials and Methods

2.1. Raw Materials

Commercial corn starch (Ingredion Ltda., São Paulo, Brazil) contained ~26–30% amylose and ~70–74% amylopectin, with <0.5% gluten and ~12% moisture. Arabic coffee husks (CH; 1.4554 ± 0.0007 g·cm⁻³) were supplied by Unique Cafés Especiais (São Lourenço, MG, Brazil) and knife-milled to <2 mm. Mineral fillers were donated by CETEM (Rio de Janeiro, Brazil): Bahia Beige marble waste (BB; 2.832 ± 0.0005 g·cm⁻³, Ourolândia, Brazil) providing Ca, and a potassium feldspar (PF; 2.8608 ± 0.0015 g·cm⁻³, Serra Negra, SE, Ourolândia, Brazil) supplying K; both minerals were sieved to <20 μm.

2.2. Preparation of Thermoplastic Starch and Biocomposites

Figure 1 schematically illustrates the processing route for the TPS/CH/PF/BB composites and the associated characterization workflow. All raw materials were dried in a forced-air oven at 60 °C to constant mass (24 h) and then stored in a desiccator for an additional 24 h. Thermoplastic starch (TPS) was prepared by premixing corn starch and glycerol at a 70/30 wt.% ratio until homogeneous. The TPS was subsequently blended with coffee husks (CH), potassium feldspar (PF), and Bahia Beige marble (BB) according to the formulations in

Table 1. Formulations based on TPS, coffee husks, potassium and Bahia Beige for extrusion.

Sample (TPS/CH/PF/BB)	TPS (%m/m)	Coffee Husks (% m/m)	Potassium Feldspar (PF) (% m/m)	Bahia Beige (BB) (% m/m)
TPS (control)	100	-	-	-
90/10/0/0	90	10	-	-
90/0/5/5	90	0	5	5
80/10/5/5	80	10	5	5
70/20/5/5	70	20	5	5

to obtain TPS/CH/PF/BB mixtures. Each formulation was compounded in a single-screw extruder (AX Plásticos, Diadema, Brazil) operating at 47 rpm, with three heating zones set to 85, 90, and 95 °C (feed-to-die). The extrudate was pelletized. Test specimens were produced by compression molding 10 g of pellets: hot pressing at 100 °C for 300 s under 6 t, followed by cold pressing for 60 s under 6 t. Unless otherwise stated, all percentages are by mass.

Processing parameters such as screw diameter, compression ratio, and cooling rate significantly influence the structure and properties of the TPS/CH/PF/BB composites. These factors affect filler dispersion, matrix homogeneity, interfacial bonding, and crystallinity. The chosen extrusion and molding conditions were optimized to ensure uniform dispersion and reproducible mechanical performance.

2.3. Characterization

2.3.1. X-Ray Fluorescence (XRF)

Mineralogical characterization of Bahia Beige, potassium feldspar, and coffee husks employed X-ray fluorescence (XRF) on an Axios Max PANalytical (WDS-2) system at 4 kW. Loss on ignition was measured with a LECO TGA-701 by applying two ramps: 25–107 °C at 10 °C min⁻¹, followed by 107–1000 °C at 40 °C min⁻¹. The procedure was terminated after three consecutive constant weight readings.

2.3.2. Fourier-Transform Infrared Spectroscopy (FTIR)

Fourier-transform infrared spectra were recorded with a Thermo Scientific Nicolet 6700. Formulations were placed on an ATR accessory fitted with a ZnSe crystal. Data acquisition used 120 co-added scans, 4.182 cm⁻¹ resolution, across 500–4000 cm⁻¹.

2.3.3. Scanning Electron Microscopy (SEM)

The SEM analysis was performed using a Hitachi TM3030Plus tabletop microscope operating at 15 kV to observe specimens coated with gold. Transverse cross-sections prepared by cryogenic fracture were examined, and micrographs were captured at 200× magnification.

2.3.4. Tensile Tests

Tensile testing followed ASTM D638 [14] on an Emic universal machine (DL-1000) equipped with a 5 kN load cell. Tests were run at 23 ± 2 °C and 50 ± 10% RH, using a crosshead speed of 10 mm min⁻¹. Type I dog-bone specimens were used, with the following dimensions according to ASTM D638: overall length of 165 mm, gauge length of 50 mm, a width of the narrow section of 13 mm, and a thickness of 3 mm.

2.3.5. Izod Impact Tests

Impact resistance was assessed by the Izod method in accordance with ASTM D256 [15]. Tests used a universal pendulum impact tester; five specimens per formulation were evaluated. Specimen dimensions were 63.5 ± 2.0 mm (length), 12.7 ± 0.2 mm (width), and 3.2 ± 0.2 mm (thickness). Each sample was mounted vertically and struck at the center with a 5.5 J pendulum.

2.3.6. Thermogravimetric Analysis (TG/DTG)

TG/DTG analyses were performed on a PerkinElmer STA 6000 with an aluminum-oxide crucible. The temperature was ramped from 30 to 500 °C at 10 °C min⁻¹ in nitrogen to evaluate thermal stability.

2.3.7. Statistical Analysis and Optimization

Optimal dosages for incorporating sustainable reinforcing agents into the Biocomposites were determined by application of a Mixture Design of Experiments. The statistical analysis was performed using StatSoft’s Statistica 14.3 Academic Ultimate software and a simplex–lattice design (3;1) was applied in this study to develop mathematical models that relate the incorporated dosages of coffee husks, potassium feldspar, and Bahia Beige marble to the mechanical and thermal properties of the Biocomposites [16].

By applying the Mixture Design of Experiments, the proportions of raw materials can be fine-tuned to achieve optimal performance and comply with technical specifications. Additionally, the behavior of the mixtures can be accurately and reliably predicted, minimizing the need for excessive materials, time, and manpower [17].

To analyze the results, the Scheffé linear regression model was applied. This model is widely used in mixture experiments where the response variable is assumed to depend on the relative proportions of the components, rather than their absolute quantities. The linear form of the Scheffé model for three components is given by Equation (1):

Y = β₁*x₁ + β₂*x₂ + β_3×3

(1)

where Y is the response variable, x₁, x₂ and x₃ represent the proportions of TPS, CH, and PF+BB, respectively, and β₁, β₂ and β₃ are the estimated model coefficients.

This modeling approach is particularly advantageous in the context of sustainable material development, as it allows for a quantitative prediction of performance based on formulation, reduces the number of required experiments, and facilitates the identification of optimal compositions with improved mechanical behavior. Furthermore, it provides a statistically robust framework for exploring synergistic or antagonistic effects among the mixture of components.

The Mixture Design resulted in five mixtures, replicated five times, totaling 25 experiments, each one performed randomly to reduce systematic errors. The composition of the mixtures, expressed in proportions, are shown in

Table 2. Component proportioning for each mixture.

Mixtures *	TPS (% m/m)	CH (% m/m)	PF+BB (% m/m)
1	100	0	0
2	90	10	0
3	90	0	10
4	80	10	10
5	70	20	10

* Five replicates of each mixture were performed, totaling 25 experiments.

. Although the Biocomposites are composed of four components: TPS, coffee husks (CH), potassium feldspar (PF) and Bahia Beige marble (BB), the statistical analysis will consider only three factors: TPS, CH and PF+BB. The choice for a three-component model is because the proportions of potassium feldspar and Bahia Beige marble are similar across all mixtures, generating multicollinearity problems in the statistical evaluation. Therefore, for the purposes of the Mixture Design of Experiments, a single factor will be considered, representing the sum of the proportions of these two components (PF+BB).

Analysis of Variance (ANOVA) was performed to evaluate such models for optimization. The dependent variables of regression model, usually called response variables, used to optimize the mechanical performance of Biocomposite were Impact Izod (II), tensile strength (TS) and elastic modulus (EM). A first-degree polynomial model was evaluated for each response variable.

The response surfaces of each dependent variable (II, TS and EM) were also obtained. Each response surface was obtained by StatSoft’s Statistica software and it is a graphical representation of the regression equation. It allows evaluating, in a more visual way, the influence of each component (chamotte, waste glass and sand) on the analyzed properties.

To prove the good fit of each regression model, a residual analysis was performed. To assess the assumptions underlying the analysis of variance, three diagnostic plots were generated and interpreted: the normal probability plot of residuals, the residuals versus fitted values plot, and the residuals versus run order plot. The normal probability plot was used to evaluate the normality of residuals, while the Residuals vs. Run order plot was analyzed to verify the independence of the residuals throughout the sequence of experiments. The combination of these diagnostics provided a robust evaluation of the model assumptions and the overall reliability of the regression model.

3. Results and Discussion

3.1. XRF Results

Figure 2 summarizes the XRF-derived compositions for Bahia Beige marble, potassium feldspar, and coffee husks. To improve interpretation, the data are shown on two y-axis windows: a 0–100 wt.% view for major phases and a 0–5 wt.% view for minor oxides.

Bahia Beige marble shows a clear dominance of CaO, representing nearly half of its total composition (~47 wt.%), which is consistent with its calcitic nature. In addition, moderate amounts of SiO₂ (~5 wt.%) and MgO (~2 wt.%) are present, while other oxides such as Al₂O₃, Fe₂O₃, and K₂O appear only in trace concentrations. The relatively high Loss on Ignition (LOI, ~43 wt.%) is attributed to CO₂ release during CaCO₃ calcination, reinforcing the predominance of calcium carbonate as the main mineral phase [13].

Potassium feldspar presents a markedly different profile, characterized by a high concentration of SiO₂ (62.6 wt.%) and significant levels of Al₂O₃ (14.3 wt.%), which together indicate its typical silicate framework. It is also noteworthy that K₂O is present at around 4.0 wt.%, confirming its role as the main potassium carrier. Smaller fractions of Na₂O and CaO are also observed, while other oxides remain below 2 wt.% (as shown in the lower-scale chart). This mineralogical balance reinforces the typical application of feldspar as a flux in the ceramic and glass industries, where K and Al contribute to structural modifications and melting behavior [13].

Coffee husks, in turn, exhibit a composition strongly influenced by their organic nature. The LOI reaches approximately 90 wt.%, reflecting the substantial mass loss from the thermal degradation of lignocellulosic components. Although inorganic oxides are present in much smaller proportions, they are not negligible. SiO₂ (~2 wt.%), K₂O (~4.3 wt.%), and CaO (~0.8 wt.%) stand out as the most relevant mineral fractions, while MgO, Al₂O₃, Fe₂O₃, and Na₂O appear close to trace levels (<1 wt.%). The relatively high K₂O content is particularly significant, as it highlights the agronomic potential of coffee husks as a nutrient source when applied to soil [13,18].

When both charts are examined together, a complementary picture emerges: the large-scale chart identifies the dominant phases (CaO in marble, SiO₂ and Al₂O₃ in feldspar, and LOI in coffee husks), while the small-scale chart reveals minor oxides that would otherwise be masked. These subtle contributions—such as Fe₂O₃ and TiO₂ in feldspar or MgO in marble—are important because, even in small amounts, they can influence color, thermal stability, and reactivity when these materials are incorporated into biocomposites.

The comparative analysis highlights the distinct chemical roles of each raw material: Bahia Beige as a calcium-rich carbonate, potassium feldspar as a silicate with alkaline potential, and coffee husks as an organic residue enriched with trace minerals. This complementarity justifies their combined use as reinforcing fillers in starch-based biocomposites, where they can contribute not only to mechanical and thermal behavior but also to potential environmental benefits [13].

3.2. FTIR Results

Figure 3a displays ATR–FTIR spectra of the base materials (500–4000 cm⁻¹), encompassing the principal absorption features. For potassium feldspar, a cluster of partially overlapping bands occurs between 650 and 1000 cm⁻¹. The 1250–830 cm⁻¹ region is assigned to asymmetric Si–O stretching, with maxima at 998 cm⁻¹ and additional features at 800 and 780 cm⁻¹ (more symmetric stretching) and at 650 cm⁻¹ (bending modes) [19,20]. Bands at 2846 and 2915 cm⁻¹, observed in both potassium feldspar and Bahia Beige, correspond to CH₂/CH₃ vibrations and have been reported previously for polypropylene composites reinforced with Bahia Beige. In the marble, peaks at 987 and 1047 cm⁻¹ (C=C bending) and at 1081 cm⁻¹ (C–O stretching) are characteristic signatures [13,21].

The infrared spectrum of coffee husk aligns with literature data [22,23], displaying peaks related to caffeine, carbohydrates, and proteins. The 3600–3000 cm⁻¹ range indicates water and hydroxyl groups, with a peak at 3301 cm⁻¹, consistent with lignocellulosic fibers [24]. The peak at 2935 cm⁻¹ corresponds to lignocellulosic components in coffee husk [25]. Caffeine absorption is observed in 1731 and 1596 cm⁻¹, with the 1700–1600 cm⁻¹ range linked to chlorogenic acid content in coffee. Additional peaks at 1241, 1014, and 771 cm⁻¹ suggest the presence of sucrose (1242–1218 cm⁻¹) and arabinogalactans (1065–1020 cm⁻¹) [26,27]. The BB infrared bands align with characteristic limestone signals at 1454–1380 and 867–755 cm⁻¹ (carbonate ion stretching) [13].

Figure 3b shows the FTIR spectra of TPS and its formulations. The OH stretching bands of starch, water, and glycerol appear at 3600–3300 cm⁻¹, while peaks at 2924 and 2893 cm⁻¹ correspond to C–H, H–C–H, and C–OH stretching. The 1640–1600 cm⁻¹ band is related to water vibrations in TPS, and the 856 cm⁻¹ peak corresponds to αD-glucose bond conformations in starch [13,28,29,30,31]. The formulation bands closely resemble those of TPS, indicating that the fillers were probably incorporated into the polymer matrix.

3.3. Morphological Analysis by SEM

Figure 4 compares the fracture-surface morphologies of neat TPS and the composite formulations at 200×. Neat TPS presents a rough, granular surface with scattered microvoids, a feature typical of semicrystalline starch matrices. The small cavities are consistent with volatilization of glycerol and/or residual moisture during processing, as also indicated by FTIR trends [12,32].

For the 90/10/0/0 composite, the TPS remains the continuous phase and the CH particles appear well encapsulated, with only occasional pull-out marks. The high polymer fraction (90 wt.%) favors wetting and physical anchoring of the lignocellulosic filler, producing a comparatively uniform morphology that aligns with the FTIR-inferred interactions between TPS and CH [12,32].

In the mineral-filled 90/0/5/5 composite, angular particulate features attributable to PF/BB are visible but remain embedded in a continuous TPS phase. The interface appears largely intact at this magnification, suggesting compatibility achieved predominantly by physical mechanisms (melting, flow and mechanical interlocking of the matrix during compounding), in agreement with Chagas et al. [12].

For the hybrid systems, increasing the overall filler content differentiates the morphologies of 80/10/5/5 and 70/20/5/5. The 80/10/5/5 sample shows clearer signs of interfacial decohesion—microcrack-like discontinuities and localized gaps around particles, indicating regions of stress concentration where the rigid minerals disrupt matrix continuity. In contrast, 70/20/5/5 exhibits a more continuous surface with fewer pronounced gaps at this scale, suggesting improved packing and TPS–CH interfacial contact despite the higher CH loading. This trend is consistent with the FTIR evidence for intermolecular interactions (hydrogen bonding) between TPS and lignocellulosic groups, which may mitigate debonding even in the presence of mineral particulates [12,32].

The delamination zones observed in the SEM images arise from local stress concentrations and differences in stiffness between fillers and the TPS matrix. Microcracks or pull-out features occur where interfacial adhesion is weaker, while good wetting and hydrogen bonding around CH particles help maintain matrix continuity. These zones reflect the balance between filler geometry, distribution, and matrix–filler interactions, influencing crack initiation and propagation. Overall, the SEM observations corroborate the spectroscopic indication of TPS–filler interactions and highlight the role of matrix continuity and filler geometry (fibrous/irregular CH vs. angular PF/BB) in governing interfacial integrity and potential crack initiation sites [12,32].

3.4. Mechanical Properties

Table 3. Summary results of mechanical properties of composite formulations.

Formulation Groups	Tensile Strength (MPa)	Strain at Break (%)	Elastic Modulus (MPa)	Impact Resistance (KJ/m2)
TPS	1.389 ± 0.066	9.818 ± 2.767	27.95 ± 5.93	13.612 ± 2.269
90/10/0/0	1.296 ± 0.065	6.361 ± 0.568	40.59 ± 0.74	20.827 ± 7.045
90/0/5/5	1.375 ± 0.559	9.238 ± 2.490	31.27 ± 3.73	15.397 ± 3.256
80/10/5/5	1.428 ± 0.218	6.737 ± 0.312	45.39 ± 5.56	16.170 ± 3.215
70/20/5/5	2.007 ± 0.148	7.462 ± 0.803	70.24 ± 5.70	14.170 ± 4.315

compiles tensile strength, strain at break, elastic modulus, and Izod impact energy for neat TPS and the filled systems. Taken together, the data reveal a progressive stiffening with filler addition, accompanied by the typical reduction in ductility; strength remains near the TPS baseline for low loadings but rises markedly only at the highest hybrid content, reflecting the usual stiffness–toughness trade-off in filled thermoplastics [33].

Starting from the TPS reference, adding a single filler causes little change in strength: −6.7% for CH (90/10/0/0) and −1.0% for PF+BB (90/0/5/5). In both cases, the creation of matrix–filler interfaces limit plastic flow and explains the concurrent drops in strain at break (−35.2% and −5.9%, respectively) [33]. When both fillers are present, the response depends on total loading: the hybrid 80/10/5/5 shows only a marginal gain in strength (+2.8% vs. TPS), whereas increasing CH in 70/20/5/5 yields a substantial rise to 2.007 ± 0.148 MPa (+44.5%). This improvement suggests more efficient stress transfer from the matrix to the combined lignocellulosic/mineral network once a critical filler fraction is reached, in line with the physical interactions promoted during processing [12].

The evolution of stiffness is more monotonic. Relative to TPS, the modulus increases by +11.9% with minerals alone (90/0/5/5) and by +45.2% with CH alone (90/10/0/0). Hybridization further amplifies stiffness to +62.4% (80/10/5/5), peaking at 70.24 ± 5.70 MPa (+151.3%) for 70/20/5/5. The trend reflects the higher rigidity of both CH and the angular mineral particulates and their growing contribution to elastic load-bearing as the matrix continuity is increasingly constrained [33].

Ductility tracks inversely with stiffness, but the extent of loss depends on formulation. All composites elongate less than TPS (−31.4% to −35.2% for 80/10/5/5 and 90/10/0/0; −24.0% for 70/20/5/5). The strongest composite (70/20/5/5) retains higher strain at break than CH-only (7.46% vs. 6.36%), indicating that hybridization with minerals can temper the loss of extensibility even at elevated filler contents [12,33].

Impact performance highlights the distinct role of CH. CH alone produces the largest gain over TPS (+53.0% in 90/10/0/0), consistent with energy-dissipating mechanisms such as microcrack deflection and particle pull-out in lignocellulosic systems [12]. Minerals alone give a modest increase (+13.1%), while hybrids fall between these limits (+18.8% for 80/10/5/5 and +4.1% for 70/20/5/5), reflecting the balance between toughening by CH and embrittlement tendencies introduced by rigid mineral particulates [12,33].

In summary, moving from TPS to the composites reveals a tunable landscape: CH prioritizes impact energy at the cost of ductility and with little strength change, minerals primarily raise stiffness with modest toughness benefits, and the most filled hybrid maximizes strength and stiffness while keeping impact near baseline and ductility at acceptable, though reduced levels.

3.5. Thermal Analysis

Figure 5 presents representative TGA and DTG curves for neat TPS and the composites. All formulations exhibit a multistep profile typical of starch-based systems: (i) a small initial mass loss below ~120 °C associated with moisture/plasticizer volatilization; (ii) the main decomposition step between ~280–330 °C, attributed to TPS depolymerization and glycosidic bond scission; and (iii) a slow evolution of residue at higher temperatures. This sequence reproduces the behavior reported for starches with different amylose/amylopectin ratios [34].

The quantitative data summarized in

Table 4. Thermal results in TG/DTG analysis.

Sample	Tonset (°C)	Tmax (°C)	Tendset (°C)	Residue (%)
TPS	263.25	312.24	334.37	0
90/0/5/5	261.83	317.02	347.12	19.08
80/10/5/5	236.45	296.12	338.14	19.88
70/20/5/5	233.07	292.35	341.66	23.30
90/10/0/0	238.17	298.60	343.01	14.75

allow for a more detailed comparison among formulations. The onset temperature (T_onset) of degradation for neat TPS was 263.25 °C, slightly higher than that of the filled systems, which ranged from 233.07 °C to 261.83 °C. This reduction in T_onset for the composites indicates that filler incorporation—particularly at higher loadings, may promote earlier thermal activation, possibly due to minor catalytic effects or enhanced heat conduction through the inorganic phase. Nonetheless, the main degradation temperature (T_max) remains close to 300 °C for all formulations, confirming that the thermal decomposition pathway continues to be governed by the TPS matrix.

A closer examination reveals that T_max is highest for 90/0/5/5 (317.02 °C) and lowest for 70/20/5/5 (292.35 °C). The hybrid compositions, especially 70/20/5/5, exhibit a broader DTG peak, suggesting a reduction in the maximum degradation rate and a more gradual mass-loss process. This broadening is consistent with the restricted chain mobility and partial heat shielding provided by the dispersed fillers, which act as physical barriers that slow down volatile release and thermal diffusion.

The endset temperature (T_endset) varies modestly among the samples (334–347 °C), indicating that degradation completion is only slightly delayed in the composites. The most significant differences, however, appear in the residual mass at 600 °C, which increases proportionally to the filler content. While neat TPS leaves virtually no residue, the composites retain between 14.75% and 23.30%, following the order 70/20/5/5 > 80/10/5/5 > 90/0/5/5 > 90/10/0/0 > TPS. This trend confirms the progressive contribution of inorganic matter (PF and BB) to the thermal stability of the system.

Overall, the fillers do not modify the primary decomposition mechanism of the starch matrix, but they subtly reduce the degradation rate and enhance the residual thermal stability. The hybrid composites, particularly 70/20/5/5, exhibit the most evident stabilization effect, attributed to the combined barrier and heat-transfer moderation induced by mineral fillers. These results align with the thermal behavior reported for other starch-based biocomposites, reinforcing that while the matrix dictates the degradation pathway, the fillers act as physical stabilizers that modulate the kinetics of mass loss and promote char formation at high temperatures [34,35].

3.6. Statistical Analysis

Figure 6 shows the statistical distribution of the response variables for the five mixtures and their replicates using box plots. While the mean values indicate the overall performance trends, the box plots reveal the variability among replicates, showing that the formulations generally exhibited narrow interquartile ranges, which indicates reproducible mechanical behavior.

ANOVA was applied to identify which factors drive the mechanical behavior of the composites. The p-values associated with TPS, CH, and PF+BB (and their interactions) indicate their influence on the responses II, TS, and EM. With a 95% confidence level (α = 0.05), factors/interactions with p ≤ 0.05 are deemed significant, whereas p > 0.05 indicates no evidence of association.

Table 5. Scheffé models obtained by statistical analysis and respective squared-R.

Response	Scheffé Models	R2	R2adjust
Impact Izod	II (kJ/m2) = 0.157 × TPS + 0.087 × CH + 0.273 × (FP+BB)	94.23%	87.67%
Elastic Modulus	EM (MPa) = 0.252 × TPS + 0.796 × CH + 0.421 × (FP+BB)	89.37%	88.41%
Tensile Strength	TS (MPa) = 0.012 × TPS + 0.036 × CH + 0.027 × (FP+BB)	93.66%	88.54%

reports the Scheffé model fitted for each response along with R² and adjusted R². R² reflects how well the model matches the data and typically rises as predictors are added. Adjusted R² accounts for model size, enabling comparison among regressions with different numbers of terms and indicating whether adding predictors improves explanatory power without overfitting.

In all the regression analyses, the three factors analyzed were significant, and the values of R² adj indicate that the final models contain terms that effectively explain the experimental data and can predict new responses with greater accuracy.

The regression models showed high coefficients of determination, with adjusted R² values above 87% for all responses. These values indicate strong model fits with minimal overfitting. All variables (TPS, CH, and PF+BB) were statistically significant (p ≤ 0.05), highlighting their substantial influence on the mechanical behavior of the composites.

To validate the assumptions of the regression models to the mechanical properties of the Biocomposites, diagnostic plots were examined for the Izod Impact, Elastic Modulus and Tensile Strength responses (Figure 7). These diagnostic plots included normal probability plots of the studentized residuals and plots of studentized residuals versus run order.

A Normal probability plot of the residuals can reflect the reliability of the data; the closer the data is to a straight line, the closer it is to a normal distribution. The normal probability plots revealed that the studentized residuals for all three models followed a reasonably linear trend, closely aligned with the expected normal distribution. This indicates that the assumption of residual normality was met, supporting the validity of the ANOVA results and the reliability of statistical inferences derived from the models.

The studentized residuals vs. run order plots demonstrated that the residuals were randomly scattered around zero, without discernible patterns, trends, or clustering over the course of the experimental sequence. This confirms that the residuals are independent and that no systematic errors or external influences affected the experimental process during data collection. Both diagnostic plots confirm that the fitted models satisfy the key assumptions of normality, homoscedasticity, and independence of residuals. Therefore, the regression models presented in Table 5 adequately describe the influence of TPS, coffee husks, and mineral fillers on the mechanical properties of the Biocomposites and are statistically robust for predictive and optimization purposes.

The response surface of Impact Izod model, illustrated in Figure 8, shows a positive correlation between all three components and impact resistance.

The PF+BB component exhibited the most significant contribution to impact resistance, followed by TPS and CH. This suggests that mineral fillers significantly enhance the material’s capacity to absorb impact energy, likely due to their contribution to stiffness, microstructural integrity, improved interface interaction and energy dissipation capacity through plastic deformation. The surface reveals a peak in impact resistance at intermediate proportions of PF+BB, confirming the optimal synergy between matrix flexibility and mineral reinforcement.

Coffee husks, although less rigid, contribute to energy dissipation through their fibrous structure and possibly increased interfacial interactions. TPS, serving primarily as the matrix, provides baseline ductility but has the least impact on improving this property.

The surface plot (Figure 9) shows a sharp rise in Elastic modulus with increasing CH content, while TPS contributes the least. This result is consistent with the SEM analysis, where increased CH promoted structural continuity and reduced porosity.

Although all factors positively influence Tensile Strength, showed in Figure 10. CH again shows the highest contribution, followed by PF+BB. The plots indicate a synergy between CH and minerals, where hybrid compositions, especially the 70/20/5/5 formulation, maximize tensile performance, as corroborated by experimental results in Table 3. Similar results were obtained by Collazo-Bigliardi [8] and Collazo-Bigliardi et al. [9], who studied composites of cellulosic fibers from coffee and rice husks incorporated into thermoplastic starch films. Cellulosic fibers from both residues were effective as reinforcing agents in films containing liquid starch.

3.7. Material Selection for Sustainable Applications

Figure 11 presents the Ashby diagram showing the relationship between tensile strength and Young’s modulus. The analysis of starch-based thermoplastic composites (TPS) in the Ashby chart reveals that all developed materials fall within the range corresponding to low-stiffness polymers and polymeric foams, characterized by a Young’s modulus between 0.028 and 0.070 GPa and a tensile strength between 1.30 and 2.01 MPa.

This region of the diagram is typically occupied by materials such as EVA, LDPE, and polyurethane foams, which combine high deformability with low elastic modulus—properties that are desirable in applications requiring flexibility, low weight, and energy absorption capability.

Among the evaluated formulations, there is a clear trend of simultaneous increase in stiffness and tensile strength as the composition approaches 70/20/5/5, which exhibits the highest modulus (70.24 MPa) and tensile strength (2.01 MPa). This behavior indicates improved cohesion within the polymeric matrix and a possible reinforcement of interfacial interactions, positioning this composite at the upper boundary of the flexible polymer region. On the other hand, the 90/10/0/0 formulation stands out for its high impact resistance (20.83 kJ/m²), demonstrating superior energy absorption performance, albeit at the expense of lower stiffness. This duality suggests the potential for targeted optimization depending on the functional requirements of the application—whether prioritizing moderate structural rigidity or higher toughness.

From the perspective of material selection for sustainable applications, the developed composites exhibit a promising profile. The use of starch as the polymer matrix provides renewability and potential compostability, significantly reducing the carbon footprint and dependence on petrochemical resources. Furthermore, the low-temperature processing typical of starch-based biopolymers contributes to reduced energy consumption during manufacturing [36]. Although the obtained mechanical properties are modest compared to those of engineering polymers, they are well-suited for short-lifespan and low-load products such as biodegradable agricultural films, flexible packaging, protective interlayers, and cushioning foams [36,37].

Nevertheless, low stiffness and moisture sensitivity remain intrinsic limitations of these materials. Sustainable enhancement strategies may include the incorporation of lignocellulosic reinforcements, nanoclays, or natural fibers derived from agro-industrial residues to increase stiffness without compromising biodegradability [38]. In parallel, applying hydrophobic coatings or developing blends with biodegradable polyesters, such as PLA or PBS, can expand the range of potential applications and improve functional durability [39,40].

4. Conclusions

In this paper, thermoplastic starch (TPS) biocomposites reinforced with coffee husk (CH) and mineral fillers: potassium feldspar (PF) and Bahia Beige marble (BB); were produced and characterized, forming hybrid composites in different proportions defined by a simplex–lattice mixture design. The TPS/CH/PF/BB formulations were processed by single-screw extrusion followed by hot/cold pressing to obtain test specimens. The reinforcements’ composition was assessed by XRF, and in the composites, we investigated chemical structure, fracture morphology, thermal stability, and mechanical properties. The results were analyzed using Scheffé polynomial models to isolate the contribution of each component and support performance optimization.

Microstructural results indicated good processability and dispersion of the reinforcements within the TPS matrix. SEM micrographs showed neat TPS with a typical rough surface and, in the composites, effective particle encapsulation, especially for 90/10/0/0 and 90/0/5/5. In the hybrid composites, localized decohesion zones were more evident in 80/10/5/5 than in 70/20/5/5, suggesting that filler type and loading govern matrix continuity and potential crack-initiation sites, in line with the physicochemical interactions discussed by FTIR.

For mechanical properties, the classical stiffness–toughness trade-off was observed. Compared with TPS (1.389 MPa; 27.95 MPa), single-filler compositions kept tensile strength close to the baseline (−6.7% in 90/10/0/0; −1.0% in 90/0/5/5), whereas the hybrids progressively increased strength, culminating in 70/20/5/5 (2.007 MPa; +44.5%). The elastic modulus rose monotonically with filler fraction, reaching 70.24 MPa in 70/20/5/5 (+151%). Ductility decreased in all compositions, but the strongest hybrid (70/20/5/5) retained higher elongation than the CH-only system, indicating that mineralization moderates the loss of extensibility. For impact toughness, CH alone delivered the largest gain (+53% in 90/10/0/0), minerals produced modest increases, and hybrids lay in between—reflecting the balance between crack-deflection/pull-out mechanisms (CH) and stiffness enhancement (PF/BB).

TGA/DTG confirmed that the degradation mechanism is dominated by the TPS matrix: all formulations showed a main step between ~280–330 °C with a DTG peak near 300 °C. Filler addition broadened and attenuated the DTG peak and, in hybrids, slightly shifted the minimum to higher temperatures, indicating a small reduction in the maximum degradation rate and a modest improvement in thermal stability. Final residue increased with the inorganic fraction (PF/BB), whereas neat TPS left only trace residue.

XRF analysis helped explain the observed behaviors: PF, rich in SiO₂/Al₂O₃ and alkalis, and calcitic BB (high CaO, high LOI) contributed to stiffness and higher thermal residue, while CH, with high LOI and predominant K₂O, favored energy-dissipation mechanisms and higher impact resistance.

Taken together, the results show that performance can be tuned by balancing CH/PF/BB: prioritizing impact and formability with higher CH or maximizing stiffness and strength with higher total filler and mineral presence. This design flexibility, combined with a scalable extrusion/pressing route, makes TPS/CH/PF/BB composites promising for low- to medium-duty applications. From a sustainability perspective, these composites valorize agro-industrial waste (coffee husk) and mineral co-products (feldspar and marble), promoting upcycling, waste reduction, and partial substitution of fossil polymers with a renewable matrix. By combining abundant, low-cost sources with a smaller carbon footprint, the proposed system aligns with circular-economy principles and paves the way for viable technical solutions in packaging, agricultural inputs, and disposable items with improved performance and reduced environmental impact.