Compaction Characteristics and Mechanical Response of Gravel–Glass–Rubber Mixtures ^†

Abstract

From a geotechnical engineering viewpoint, recycling and reuse of crushed glass and tire rubber can significantly help reduce the demand for natural resources (i.e., sand and gravel aggregates). Following an earlier study by the authors aimed at characterizing gravel–rubber mixtures (GRM), this paper focuses on the geotechnical assessment of gravel–glass–rubber mixtures (GGRM) made of recycled crushed green glass bottles and recycled granulated tire rubber. Specifically, the compaction, one-dimensional compressibility, and shear strength characteristics of GGRM prepared at 40% and 55% rubber content by volume (R_B) with varying glass content by volume (G_L) are investigated. It is found that compacted GGRM possesses high strength (i.e., friction angle ≥ 30°) and adequate compressibility, making it a suitable general and structural fill material for use in eco-friendly geotechnical applications.

1. Introduction

Traditionally, freshly quarried aggregates and dredged soils have been the primary source of construction materials in geotechnical applications. However, nowadays, the recycling of waste materials (i.e., industrial by-products, commercial waste, and construction and demolition rubble) is becoming a priority. This shift towards secondary usage of waste materials has numerous benefits, such as reducing the need for fresh primary materials, minimizing waste disposal, benefiting the environment, enhancing sustainability, reducing carbon emissions, and promoting a circular economy [1,2,3,4,5,6].

In New Zealand, each year, over 6.3 million waste tires are produced [7]; additionally, 120,000 tons of glass and 380,000 tons of plastic are sent to landfills [8]. The amount of waste sent to landfills increased by 47% from 2009–10 to 2018–19 [8]. The large volume of discarded materials highlights the need to reuse and recycle waste materials [2], which is a crucial strategic approach for meeting the UN Sustainable Development Goals (SDGs) [9] and becoming a more sustainable country. From a geotechnical engineering perspective, recycled crushed glass [10,11,12,13,14,15,16,17,18,19] and recycled tire rubber inclusions [20,21,22,23,24,25,26,27,28,29,30,31,32] have excellent mechanical properties and durability. These waste materials are readily available and cost-effective. The benefits of using such recycled materials are significant, especially if they are used as a replacement for fresh construction materials made from non-renewable resources, e.g., quarried gravel and sand. The engineering characteristics of discarded granular materials can be utilized in a beneficial way to promote sustainability in the geotechnical construction sector [6].

The reuse of granulated rubber derived from waste on its own or mixed with granular soils has been found suitable in applications such as lightweight embankment fill, drainage layers, conventional fill, and retaining wall backfill [20,21,22,23,24,25,26,27,28,29,30,31,32]. Direct shear strength parameters reported in the literature for pure rubber, sand–rubber mixtures (SRM), and gravel–rubber mixtures (GRM) have been comprehensively summarized by Tasalloti et al. [33]. Although the friction angle (15–35°) of pure rubber is generally lower than pure soils, there is a range of volumetric rubber content (R_B) (i.e., 20% < R_B < 50%) within which the friction angle of SRM exceeds that of the pure sand. This means that the shear strength of pure sand can be enhanced by mixing it with rubber. For these mixtures, the friction angle varies between 26° and 62°. Compared with SRM, tests on GRM are still very limited. Yet, the available data suggest that the friction angle of GRM generally decreases with increasing R_B, so the addition of rubber does not improve the strength of gravel soils. Yet, it can be debated that the addition of gravel can enhance the mechanical properties of highly compressible and low-strength rubber as compared with sand.

In the New Zealand context, the recycling of waste tire-derived granulated rubber mixed with gravel for use in eco-friendly geotechnical projects has been investigated by Chiaro et al. [34,35,36], Tasalloti et al. [33,37,38,39], and Banasiak et al. [40,41]. It has been found that GRM with R_B ≤ 40% have adequate strength, low compressibility, and high energy dissipation properties; hence, they can be used as structural fill materials. GRM with 40% < R_B ≤ 55% may also be appropriate as fill materials if their compressibility meets performance requirements. The use of gravel-size rubber inclusions in GRM has also been recommended. While different rubber sizes show minimal to no effect on the peak shear strength, the larger rubber inclusions (similar in size to the gravel grains) show less compressibility [33,34,35,36,37,38,39] and the least leaching of toxic chemicals [40,41].

Research into the geotechnical engineering characteristics of crushed glass alone has been carried out by several researchers [10,11]. Results of direct shear tests indicate that the friction angle is 47°–62° [10], which is comparable to that of pure gravel, i.e., 54°–61° [37,38]. Thus, recycled glass could potentially be used in geotechnical applications to partially replace gravel in GRM and produce gravel-recycled glass-recycled rubber mixtures (GGRM).

With the main purpose of developing more eco-friendly geotechnical construction materials by reusing crushed glass and recycled tire rubber as a partial substitute for freshly quarried gravel, this study aims to examine the compaction and mechanical response (i.e., compressibility and strength) of GGRM by means of detailed laboratory investigations, including standard Proctor compaction, one-dimensional compression, and direct shear tests, building on the previous studies on GRM carried out in the geotechnical laboratory of the University of Canterbury, New Zealand, and reported by Chiaro et al. [34,35,36], Tasalloti et al. [33,37,38,39], and Banasiak et al. [40,41].

2. Materials

The materials assessed in this study consist of pea gravel (specific gravity, G_S = 2.66; mean diameter D₅₀ ≈ 5.5 mm), granulated recycled rubber (G_S = 1.15; D₅₀ ≈ 5.2 mm), and crushed green glass (G_S = 2.50; D₅₀ ≈ 5.5 mm). The gravel and rubber were both commercially sourced. The gravel was washed and dried before any testing began to remove any trace of fine material. The rubber is free from steel wires and textile reinforcement. The crushed green glass consisted of a range of angular particle shapes and sizes, with a few long, slender pieces. It was obtained from wine bottles that were stripped of their labels and cleaned before crushing.

Previous studies on GRM have shown that the use of smaller rubber inclusions produces materials with lower friction angles and high compressibility [38]. Additionally, other studies reported in the literature have demonstrated that sand mixed with large rubber mixtures also produces highly compressible materials [29]. Thus, the authors have recommended the use of gravel particles and rubber inclusions of the same D₅₀ or aspect ratio = 1 to form GRM [38]. In the case of crushed glass fragments, preliminary investigations, propaedeutic to the GGRM reported in this study, have indicated that the use of larger glass results in greater crushability of the GGRM. Therefore, to minimize the crushability of large glass fragments and the compressibility of large rubber inclusions, rubber and glass having the same size as the gravel were used in this study to form GGRM.

Representative photos of the various materials are shown in Figure 1a, along with their particle size distribution (PSD) curves (Figure 1b). As schematically shown in Figure 1b, the gravel has rounded to subrounded grains with aspect ratios (H/L) ranging from 1 to 1.6, the glass fragments have a plate-like shape of 2.8 mm thickness on average and aspect ratios varying between 1 and 1.8, and the rubber inclusions have a cube-like shape with aspect ratios in the range of 1 to 1.5 and an average thickness of 5 mm.

As shown in Figure 2, different GGRM mixtures were formed by keeping the R_B constant at 0.4 and 0.55 (i.e., 40% and 55%) but changing the proportion by volume of glass (G_L) and gravel (G_R) in the mixtures.

The volumetric fraction of each material was calculated using Equations (1)–(3):

$$G_{\mathrm{L}}=V_{\mathrm{G}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}}/(V_{\mathrm{G}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}}+V_{\mathrm{G}\mathrm{r}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{l}}+V_{\mathrm{R}\mathrm{u}\mathrm{b}\mathrm{b}\mathrm{e}\mathrm{r}})$$

(1)

$$R_{\mathrm{B}}=V_{\mathrm{R}\mathrm{u}\mathrm{b}\mathrm{b}\mathrm{e}\mathrm{r}}/(V_{\mathrm{G}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}}+V_{\mathrm{G}\mathrm{r}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{l}}+V_{\mathrm{R}\mathrm{u}\mathrm{b}\mathrm{b}\mathrm{e}\mathrm{r}})$$

(2)

$$G_{\mathrm{R}}=V_{\mathrm{G}\mathrm{r}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{l}}/(V_{\mathrm{G}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}}+V_{\mathrm{G}\mathrm{r}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{l}}+V_{\mathrm{R}\mathrm{u}\mathrm{b}\mathrm{b}\mathrm{e}\mathrm{r}})$$

(3)

where V_Glass, V_Rubber, and V_Gravel are the volumes of the glass, rubber and gravel in the mixture, respectively.

3. Experimental Procedure

3.1. Compaction

To guarantee the satisfactory performance of any geotechnical structure, it is necessary to appropriately control the compaction and correctly evaluate the physical properties of the compacted materials. For GRM, Tasalloti et al. [38] reported that the Proctor compaction test is a more suitable testing procedure than the vibratory table test due mainly to the low moduli and ability of the rubber aggregates to absorb vibration energy.

Hence, in this study, the compaction characteristics of GGRM were determined by means of standard Proctor tests [42]. The steel mold used in the investigation had a diameter of 152.7 mm and a height of 116.4 mm. In each test, 3 layers were compacted by applying 56 blows of impact load. To streamline the process and maintain consistency, an automatic compaction device was used. To limit the segregation between the different materials, the mixtures were prepared in batches of approximately 900 g and then poured into the compaction mold very carefully.

Preliminary tests indicated that for specimens with a water content of 5% or more, water was lost through the base of the mold. This is due to the high permeability (i.e., free-draining) nature of GGRM. Thus, being unable to control the water content, all further compaction tests were conducted on dry specimens.

3.2. One-Dimensional Compression

Material compressibility is one of the most important factors required in design considerations. For gravels with rigid particles, any change in volume is due to the rotation, movement, and rearrangement of non-compressible grains [43]. The compressibility of GGRM, however, is entirely different from that of hard-grain gravels due to significant differences in the elastic modulus of the rigid gravel grains, the rigid but crushable glass, and the soft rubber.

The compression behavior of granular media is usually studied in conventional one-dimensional (1D) compression equipment [44]. Yet, practical difficulties are encountered when testing soils with large particles using conventional small-size compression apparatus. Therefore, in this study, the 1D compression behavior of GGRM was evaluated using an in-house-built medium-size compression cell, which can host specimens of 250 mm in inner diameter and up to 100 mm in height.

Dry specimens were prepared by the tamping method at a dry density corresponding to 90–95% degree of compaction. In the tests, the nominal vertical stress varied incrementally from 0 to 200 kPa according to the loading pathway shown in Figure 3. At the end of each loading–unloading–reloading phase, the vertical effective stress was kept constant (creep) until the settlement increment became negligible (i.e., Δs ≤ 0.02 mm/5 min). The measured vertical stress was corrected for the loss of transferred load (due mainly to material–wall friction) that was experimentally measured by using a load cell placed on the base plate of the compression device.

3.3. Shear Strength

Shear strength is another key characteristic contributing to the performance of geotechnical structures. The shear strength is the result of friction and interlocking between particles and bonding at particle contacts. Due to interlocking, GGRM may contract or expand in volume as they are subjected to shear strains.

Direct shear box tests were conducted (in accordance with the ISO Standard [45]) on dry specimens prepared by the tamping method at 90–95% degree of compaction. The size of the soil box was 100 mm (width) × 100 mm (length) × 53 mm (height). The specimens were tested under 30, 60, and 100 kPa effective normal stress (σn′)—the latter one corresponding to the maximum load capacity of the device. After the normal stress was applied to the mixtures, sufficient time was allowed for the material to fully compress. The 100 kPa tests required the most time to consolidate, approximately two hours. Once the vertical settlement increment became negligible, the specimens were sheared at a constant horizontal displacement rate of 1 mm/min. Tests were concluded when the horizontal displacement (ΔH) achieved 15–20 mm.

It is important to mention here that, while σ_n′ is limited to 100 kPa, it represents typical field stress conditions where GGRM can be used primarily as lightweight backfill materials, shallow underground layers for mitigation of liquefaction phenomena, and geotechnical seismic isolation for low-to-medium-rise lightweight structures/infrastructure [46,47,48,49,50,51,52,53]. Testing GGRM under higher confining pressures was out of the scope of this study; nevertheless, additional tests may be necessary in the future when dealing with specific applications where higher normal stresses are envisioned.

4. Experimental Results and Discussions

4.1. Compaction Characteristics of GGRM

The results of compaction tests are reported in Figure 4a and

Table 1. Results of standard Proctor compaction tests carried out on dry specimens of GGRM.

Materials			Dry Density (kg/m3)		References
Gravel GR	Rubber RB	Glass GL	Measured #	Prediction Equation (4)
1	-	-	1754	1754	Tasalloti et al. [38]
0.9	0.1	-	1621	1643
0.8	0.2	-	1514	1533
0.75	0.25	-	1487	1477
0.6	0.4	-	1315	1312
0.55	0.45	-	1248	1257
-	1	-	649	649
0.6	0.4	-	1310	1310	This study
0.45	0.4	0.15	1274	1277
0.3	0.4	0.3	1232	1243
0.15	0.4	0.45	1223	1210
-	0.4	0.6	1167	1176
0.45	0.55	-	1141	1146
0.3	0.55	0.15	1132	1112
0.15	0.55	0.3	1095	1079
-	0.55	0.45	1058	1045
1	-	-	1749	1749
-	1	-	652	652
-	-	1	1526	1526

^# Standard compaction tests on dry specimens (average values obtained from three tests).

for all tested materials. Test results for GRM reported by Tasalloti et al. [38], who used the same gravel and rubber materials used in this study, are also reported for the sake of comparison. It is observed that the dry density (ρ_dry) of GGRM decreases linearly with increasing G_L from 1310 kg/m³ to 1167 kg/m³ for R_B = 0.4 and from 1141 kg/m³ to 1058 kg/m³ for R_B = 0.55. This is mainly due to the slightly different specific gravity of the gravel (G_S = 2.66) and of the green glass (G_S = 2.50). The lower values of ρ_dry obtained for GGRM (with R_B = 0.4 and 0.55), as compared with the glass (ρ_dry = 1526 kg/m³) and the gravel (ρ_dry = 1749 kg/m³) alone, are the results of the presence of lightweight rubber inclusions (G_S = 1.15; ρ_dry = 652 kg/m³) in the mixtures.

As reported in Figure 4b, test results from Tasalloti et al. [38] show that ρ_dry of GRM (i.e., G_L = 0) decreases linearly with increasing R_B; this was confirmed in this study. Moreover, this study indicates that the ρ_dry of glass–rubber mixes (i.e., G_R = 0) also decreases linearly with increasing R_B. The linear trends shown in Figure 4b were attained by using Equation (4), which is valid for GGRM. Specific equations for gravel–glass mixtures (R_B = 0), GRM (G_L = 0), and glass–rubber mixtures (G_R = 0) are reported in

Table 2. Predictive equations for the estimation of dry density of GGRM.

Mixture Type	Equation
Gravel–glass–rubber	${\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{m}\mathrm{i}\mathrm{x}\right)}={\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{l}\right)}{·G}_{\mathrm{R}}+{\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{g}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}\right)}·G_{\mathrm{L}}+{\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{r}\mathrm{u}\mathrm{b}\mathrm{b}\mathrm{e}\mathrm{r}\right)}·R_{\mathrm{B}}$	(4)
Gravel–glass (RB = 0)	${\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{m}\mathrm{i}\mathrm{x}\right)}={\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{l}\right)}·G_{\mathrm{R}}+{\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{g}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}\right)}{·G}_{\mathrm{L}}$	(4a)
Gravel–rubber (GL = 0)	${\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{m}\mathrm{i}\mathrm{x}\right)}={\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{v}\mathrm{e}\mathrm{l}\right)}·G_{\mathrm{R}}+{\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{r}\mathrm{u}\mathrm{b}\mathrm{b}\mathrm{e}\mathrm{r}\right)}{·R}_{\mathrm{B}}$	(4b)
Glass–rubber (GR = 0)	${\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{m}\mathrm{i}\mathrm{x}\right)}={\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{g}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}\right)}·G_{\mathrm{L}}+{\rho }_{\mathrm{d}\mathrm{r}\mathrm{y}\left(\mathrm{r}\mathrm{u}\mathrm{b}\mathrm{b}\mathrm{e}\mathrm{r}\right)})·R_{\mathrm{B}}$	(4c)

for completeness. Interestingly, as shown in Figure 4c, the data points for GGRM with R_B = 0.4 and 0.55 can be connected by linear trends that are parallel to those of the gravel–glass mixtures. Similarly, as shown in Figure 4d, the data points for GGRM with varying G_R = 0–0.6 are aligned with linear trends that are parallel to those of the glass–rubber mixtures.

Sieve analyses were carried out before and after each compaction test (Figure 5a) to evaluate the amount of glass breakage/crushing induced by compaction (calculated with respect to the mass of glass in the mixture); for instance, for the green glass alone (G_L = 1), the amount of crushed material passing through the 3.35 mm sieve was about 6.9% (i.e., before compaction, the amount of glass retained at the 3.35 mm sieve was zero). In the case of GGRM (G_L = 0.15–0.6), an increase in glass breakage with increasing G_L was observed (Figure 5b), with the maximum amount of breakage (2.24%) obtained for R_B = 0.4 (G_L = 0.6). This is associated with the higher amount of glass in the mixtures. In contrast, the glass breakage seems to decrease with increasing R_B in the mixture; that is, because soft rubber inclusions in the GGRM act as a cushion that absorbs part of the impact energy delivered to the specimen, the higher amount of rubber in the mixture helps reduce glass breakage during compaction.

4.2. Compression of Compacted GGRM Under One-Dimensional Loading

The normal compression (virgin loading) response of GRM and GGRM obtained from 1D compression tests with creep is reported in Figure 6 and Figure 7 (note that such normal compression curves have been extracted for the unloading/reloading compression tests that are reported in Figure 8 and Figure 9).

From Figure 6a, it is evident that the addition of rubber to gravel generates a much more deformable material (e.g., under 100 kPa vertical effective stress (σ_v′), the gravel experiences approx. 1% volumetric strain (ε_v), while those of GRM are on the order of ε_v = 9.5% and ε_v = 10.5% for R_B = 0.4 and R_B = 0.55, respectively). These test results are consistent with those reported by Tasalloti et al. [37].

Figure 6b,c indicate that the addition of green glass to GRM (R_B = 0.4 and R_B = 0.55) produces an increase in the 1D compression. Such an increase is proportional to G_L. Furthermore, GGRM with R_B = 0.4 performed better than those with R_B = 0.55 (i.e., a much smaller ε_v is developed under the same applied σ_v′). For example, in the case of G_L = 0.45, ε_v is 11.5% for R_B = 0.4 and 19.1% for R_B = 0.55, respectively.

Figure 7 shows the variation of the secant drained constrained modulus (M = σ_v′/ε_v) with σ_v′ for GGRM with R_B = 0.4 and R_B = 0.55. M of gravel is also reported for completeness. In a log–log plot, M increases linearly with increasing σ_v′; however, it decreases with increasing R_B and G_L.

The addition of rubber inclusions to gravel induces a drastic reduction in stiffness that can be associated with the low modulus and deformability of the rubber particles. The data also suggest that as G_L increases, GGRM stiffness gradually reduces, and compressibility increases, mainly due to glass crushability. Moreover, at lower G_L, the mixtures transfer the load mainly through the gravel–rubber particle-to-particle contacts, while at larger G_L, the load is transferred primarily through the glass–rubber particle-to-particle contacts. Therefore, there is a gradual transition in the material type, from hard-soft mixtures [54] to soft–crushable mixtures; that is, the more rubber and glass are present in the mixtures, the more compressible the material becomes, and thus the lower the M.

In Figure 8 and Figure 9, the unloading–reloading 1D compression responses of GGRM are reported in terms of settlement vs. time and ε_v vs. σ_v′ relationships, respectively. A sudden initial (plastic) settlement can be observed for GRM, which is the result of the rubber particle deformability and the rearrangement of non-compressible gravel grains. When glass is added to the mixtures to form GGRM, the glass grain rearrangement and crushing also play a key role, and a much larger initial settlement takes place (Figure 8). Nevertheless, if the mixtures are subjected to unloading and reloading, their compressibility is drastically reduced, and their behavior becomes essentially elastic (Figure 9). This implies that during the unloading–reloading phase, the glass crushing and gravel grain rearrangement (mainly responsible for the plastic deformation) are minimized and that the observed elastic response is due primarily to the rubber rebound.

4.3. Strength Characteristics of Compacted GGRM from Direct Shear Tests

Results of the direct shear tests showing the different shear stress–horizontal strain relationships and volumetric responses for pure gravel, crushed green glass, and granulated tire rubber under 30, 60, and 100 kPa normal stress (σ_n′) are shown in Figure 10.

The gravel displays stiff behavior typical of dense, hard-grained soils, characterized by a high peak strength (e.g., 137 kPa at σ_n′ = 100 kPa) and brittle failure. Its volumetric response is primarily dilative. In contrast, the soft rubber has a ductile response with no clear peak and contractive volumetric behavior; it has a peak strength of 54 kPa at σ_n′ = 100 kPa. The green glass strength is in between that of gravel and rubber (i.e., the peak strength is 91 kPa σ_n′ = 100 kPa), but its volumetric behavior is similar to that of stiff gravel, although much less dilative. This is because green glass particles, although stiff, are highly crushable under shearing loading, especially when subjected to higher σ_n′ levels.

The direct shear behavior of GGRM with R_B = 0.4 and 0.55 is reported in Figure 11 and Figure 12. Due to the presence of softer rubber particles, the mechanical response of the mixtures is more ductile than that observed for pure gravel and pure glass. The variation of G_L (= 0 − 0.6) in the mixtures does not affect the overall response qualitatively or quantitatively, indicating that R_B is primarily responsible for the observed ductile behavior.

Tasalloti et al. [15] conducted direct shear tests on GRM over a range of σ_n′ from 6.5 kPa to 100 kPa. Figure 13a reports the peak strength (τ_peak) determined for gravel, rubber and GRM (R_B = 0.4) using the linear Mohr–Coulomb (MC) failure criterion as described by Equation (5):

$${\tau }_{\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}}=c_a^{\prime }+{\sigma }_n^{\prime }\tan \left({\varphi }_{\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}}\right)$$

(5)

The strength is characterized by friction angles (ϕ_peak) values of 53.2° (gravel), 34.0° (GRM, R_B = 0.4), and 26.0° (rubber). Moreover, the MC method indicates that GRM may have an apparent cohesion (c_a′), whose values are 3.2 kPa (gravel), 7.2 kPa (GRM, R_B = 0.4), and 6.1 kPa (rubber). Nevertheless, unbounded materials like gravel, rubber, and GRM are not expected to have interparticle forces leading to cohesion. A similar observation has been reported by Ghaaowd et al. [55], who conducted direct shear tests on large rubber particle sizes. Hence, although c_a′ may be seen as a convenient strength parameter, it does not provide an accurate representation of the actual strength of cohesionless (c′ = 0) materials like GRM, especially at very low-stress levels.

As shown in Figure 13b, because of the cohesionless nature of GRM, a nonlinear power form fitting curve can be adopted to better capture the strength dependency of GRM with the normal stress as expressed by Equation (6):

$${\tau }_{\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}}=a{\left({\sigma }_n^{\prime }\right)}^b$$

(6)

where a and b are material fitting parameters.

Similarly, as shown in the example in Figure 14, the MC approach considering c_a′ overestimates the strength of unbounded GGRM for σ_n′ < 30 kPa. Alternatively, if c′ = 0 is imposed, then the strength is underestimated for σ_n′ < 80 kPa and overestimated for σ_n′ > 80 kPa. On the other hand, the nonlinear approach (i.e., Equation (6)) better captures the stress dependency of the strength to the normal stress. The values of the a and b coefficients obtained for all GGRMs tested in this study are reported in Figure 14 for completeness.

In this study, the evolution of the secant friction angle (ϕ_sec = tan⁻¹ (τ/σ_n′)) during shearing was also evaluated for the different σ_n′ levels. From Figure 15 and Figure 16, it is evident that ϕ_sec is stress-dependent and decreases with increasing σ_n′. Specifically, ϕ_sec(peak) decreases from 40–48° at σ_n′ = 30 kPa to 33–37° at σ_n′ = 100 kPa for R_B = 0.4, and from 40–44° at σ_n′ = 30 kPa to 30–34° at σ_n′ = 100 kPa for R_B = 0.55.

Combining the definition of ϕ_sec and Equation (6), then ϕ_sec(peak) can be expressed by the nonlinear function shown in Equation (7):

$${\varphi }_{\mathrm{s}\mathrm{e}\mathrm{c}\left(\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}\right)}={\mathrm{t}\mathrm{a}\mathrm{n}}^{-1}\left({\displaystyle \frac{{\tau }_{\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}}}{{\sigma }_n^{\prime }}}\right)={\mathrm{t}\mathrm{a}\mathrm{n}}^{-1}\left[a{\left({\sigma }_n^{\prime }\right)}^{b-1}\right]$$

(7)

The values of ϕ_sec(peak) experimentally evaluated and those estimated by Equation (7) are reported in

Table 3. Strength parameters for GGRM evaluated by different failure criteria.

Mixture		Experimental Data ϕsec(peak) (°) at σn′			Mohr–Coulomb (Equation (5))			Nonlinear Model (Equation (7)) ϕsec(peak) (°) at σn′
					(c′ = ca′)		(c′ = 0)
RB	GL	30 kPa	60 kPa	100 kPa	ϕpeak (°)	ca′ (kPa)	ϕpeak (°)	30 kPa	60 kPa	100 kPa
0.4	0	48.2	39.5	36.9	30.8	14.9	38.4	47.5	40.9	36.1
0.4	0.15	45.0	40.3	33.1	26.3	17.4	35.8	45.7	38.8	33.9
0.4	0.3	45.8	37.1	33.6	27.0	15.3	35.4	45.4	38.1	33.0
0.4	0.45	41.8	39.6	34.9	31.2	10.3	36.6	42.3	38.4	35.6
0.4	0.6	40.5	37.9	33.8	30.4	9.3	35.3	40.8	37.0	34.3
0.55	0	44.0	39.2	34.3	29.0	13.6	36.2	44.3	38.6	34.6
0.55	0.15	40.3	36.8	33.8	30.6	8.3	35.0	40.4	36.6	33.9
0.55	0.3	41.2	33.5	32.3	28.0	9.4	33.2	40.5	35.1	31.4
0.55	0.45	40.3	34.1	29.9	24.5	12.3	31.7	40.3	34.1	29.9
Mixture					Model parameters			Duncan Nonlinear Model [56]
								ϕsec(peak) (°) at σn′
RB	GL				ϕ0 (°)	Δϕ (°)		30 kPa	60 kPa	100 kPa
0.4	0				35.9	−9.6		47.6	40.9	36.0
0.4	0.15				33.8	−9.7		45.5	38.8	34.1
0.4	0.3				32.8	−10.3		45.3	38.2	32.9
0.4	0.45				35.5	−5.6		42.3	38.4	35.6
0.4	0.6				34.0	−5.5		40.7	36.9	34.1
0.55	0				34.5	−8.0		44.2	38.7	34.6
0.55	0.15				33.8	−5.4		40.4	36.6	33.9
0.55	0.3				31.2	−7.6		40.4	35.2	31.3
0.55	0.45				29.7	−8.7		40.3	34.3	29.8

for all the GGRM and the three different σ_n′ levels. The ϕ_peak values obtained by the MC method with and without c_a′ are also reported in Table 3 for comprehensive.

For the purpose of validation of the nonlinear power–law model in Equations (6) and (7), secant friction angle values obtained by using the well-known nonlinear strength model of Duncan [56] are also provided in Table 3 for all GGRM and the three levels of σ_n′ = 30, 60, and 100 kPa. The Duncan model is described by Equation (8):

$${\varphi }_{\mathrm{s}\mathrm{e}\mathrm{c}\left(\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{k}\right)}=∆\varphi \log \left({\displaystyle \frac{{\sigma }_n^{\prime }}{101.3}}\right){\varphi }_0$$

(8)

where $∆\varphi$ and ${\varphi }_0$are fitting parameters; the values obtained for GGRM are reported in Table 3. There is a very good agreement between the estimates of the two methods (i.e., the maximum difference is 0.2°), indicating the suitability of the nonlinear power–law model for the evaluation of GGRM strength.

Based on the direct shear test results and analyses reported above, the ϕ_peak values of GGRM with R_B = 0.4 and 0.55 are typically higher than 25–30°, making GGRM suitable structural fills [1] for typical geotechnical applications (e.g., embankments, foundations, and backfills for retaining structures).

4.4. Behavioral Zones for Compacted GGRM

From a mechanical viewpoint, it is well understood that for GRM, the addition of rubber inclusions to gravel induces a drastic reduction in stiffness that can be associated with the low modulus and deformability of the rubber inclusions [37,38,54]. By examining the load-transfer mechanisms of the GRM tested by Tasalloti et al. [38]—using a combination of numerical simulations (Discrete Element Method) and micro-mechanical analyses—Chew et al. [46] have identified three distinct material-like behavioral responses for GRM (Figure 17): gravel-like for 0.3 < R_B, dual (intermediate) behavior for 0.3 ≤ R_B < 0.6, and rubber-like behavior for R_B ≤ 0.6.

In stiff, gravel-like materials, the load-transfer mechanism is primarily governed by the interaction between gravel grains due to the limited amount of rubber inclusions in the mixtures; this leads to a stiff (brittle) mechanical response. Instead, in soft rubber-like materials, the load-transfer mechanism is due mainly to the interaction between rubber inclusions because of the limited amount of gravel grains in the mixtures; this leads to a soft (ductile) mechanical response. Additionally, in the case of dual (intermediate) materials, the strong force network responsible for the load-transfer mechanism is jointly shared between the gravel grains and rubber inclusions; this results in an intermediate stiff-soft (brittle/ductile) mechanical response.

The test results reported in this study indicate that the mechanical response of GGRM (R_B = 0.4 and 0.55) is similar to that of GRM (R_B = 0.4 and 0.55); thus, the investigated GGRM can be considered intermediate materials whose mechanical response is expected to be in between that of the stiff gravel, crushable glass, and soft rubber (Figure 17 and

Table 4. Behavioral zones identified in this study for compacted GGRM.

Symbol	Behavioral Type	Remarks
S	Soft	RB ≥ 0.6; rubber-like materials, the load-transfer mechanism is due primarily to the interaction between soft rubber inclusions
G	Stiff	RB < 0.3 and GL < 0.3; gravel-like materials, the load-transfer mechanism is due mainly to the interaction between stiff gravel grains
C	Crushable	RB < 0.3 and GR < 0.3; glass-like materials, the load-transfer mechanism is due mostly to the interaction between crushable glass particles
GC	Stiff–crushable	RB < 0.3, GL ≥ 0.3 and GR ≥ 0.3; intermediate materials, the load-transfer mechanism is primarily shared between gravel grains and glass particles; rubber inclusions are mostly inactive
GS	Stiff–soft	0.3 ≤ RB < 0.6 and GR > 0.4; intermediate materials, the load-transfer mechanism is principally shared between gravel grains and rubber inclusions; glass particles are mostly inactive
SC	Soft–crushable	0.3 ≤ RB < 0.6 and GL > 0.4; intermediate materials, the load-transfer mechanism is mainly shared between glass particles and rubber inclusions; gravel grains are mostly inactive
GSC	Stiff–soft–crushable	0.3 ≤ RB < 0.6, GL ≤ 0.4 and GR ≤ 0.4; intermediate materials, gravel grains, glass particles, and rubber inclusions actively and jointly contribute to the load-transfer mechanism

). Specifically, the experimental data show that as G_L increases (i.e., glass replaces gravel in the mixtures), the stiffness of the mixtures gradually reduces, and compressibility increases, mainly due to glass crushability. That is, at lower G_L, the mixtures transfer the load mainly through the gravel-rubber particle-to-particle contacts, while at larger G_L, the load is transferred primarily through the glass–rubber particle-to-particle contacts. Therefore, there is a gradual transition in the material type, from stiff–soft mixtures to soft–crushable mixtures, as glass is added in the mixtures as a replacement for gravel, as shown in Figure 17. Table 4 provides a description of the behavioral types identified for GGRM.

It is worth mentioning that the boundaries for GC and GSC, shown in Figure 17, have been defined based on the combination of macro-mechanical analyses of the GGRM behavior observed in this study and previous micro-mechanical and force chain analyses conducted by the first author for GRM and reported in Chew et al. [54]. While they capture the actual behavior expected for GGRM, their position may be slightly different from the actual one. Further investigations into the load-transfer mechanism at a particle level (i.e., strong force network) would be required to confirm the exact position of these boundaries. To do so, DEM numerical simulations and subsequent micro-mechanical analyses, like those carried out by Chew et al. [54] for GRM, are therefore advised in future studies.

5. Potential Use of GGRM as Fill Materials in Geotechnical Applications

The prospect of mixing a variety of recycled materials, such as crushed glass and granulated rubber, with gravel holds practical significance. The results of earlier investigations on GRM indicate that such synthetic materials can be used as geomaterials and primarily as lightweight backfill materials, shallow underground layers for mitigation of liquefaction phenomena, and geotechnical seismic isolation for low-to-medium-rise lightweight structures/infrastructure [33,34,35,36,37,38,39,40,41,46,47,48,49,50,51,52,53]. Due to the many similarities between GRM and GGRM, it is therefore expected that compounds made of gravel, crushed glass, and granulated rubber will also have adequate strength, compressibility, and minimal leaching aspects, making them suitable geomaterials [57].

In previous studies by Chiaro [58] and Forcellini [59], GRM has been primarily studied and optimized to be used in geotechnical seismic isolation (GSI) foundation systems to increase the seismic resistance of lightweight residential buildings. The results of dynamic cyclic triaxial tests [60], impact hammer tests on a physical model [59], and advanced finite element numerical simulations [58,59] have indicated that GRM (R_B = 0.4 and 0.55) can effectively filter the seismic waves passing through the GRM layer and reduce the seismic load transferred to the superstructure by 50% or more. Based on these findings, GS and potentially GSC mixtures may be used in GSI foundation systems.

Overall, the investigated GGRM (R_B = 0.4 and 0.55) would have adequate strength (i.e., peak friction angle of 30° or greater) to be used as fill materials in eco-friendly geotechnical projects [38]. However, because of their high compressibility at low G_R values (i.e., for high combined G_L and R_B values) and the different mechanical behavior they can exhibit under applied loads (Figure 17), their use as structural fills may not always be feasible or may require some improvement, as indicated in Figure 18.

Specifically, based on the mechanical responses observed in direct shear and one-dimensional compression tests and considering the behavioral zones described earlier in Table 4 and reported in Figure 17, the following recommendations can be made for GGRM (R_B = 0.4 and 0.55):

• GS mixtures (0.3 ≤ R_B < 0.6 and G_R > 0.4)—due to their high strength and limited compressibility, they can be used as structural fills. Yet, in situations where high vertical loads are applied, improvement by preloading or other ground improvement techniques should be considered to enhance their compressibility.
• GSC mixtures (0.3 ≤ R_B < 0.6, G_L ≤ 0.4, and G_R ≤ 0.4)—due to their compressibility and potential high crushability, they should only be employed as structural fills if improved, for instance, by preloading; alternatively, they can be safely used as general fills.
• SC mixtures (0.3 ≤ R_B < 0.6 and G_L > 0.4)—due to their higher compressibility and high crushability, their use as structural fills is discouraged unless improved. But they are acceptable as general fills.

For other GGRM materials not directly investigated in this study but of practical interest, the following recommendations can be made:

• G mixtures (R_B < 0.3 and G_L < 0.3)—they can be safely used as structural fills due to their high frictional resistance and minimal compressibility [38].
• GC mixtures (R_B < 0.3, G_L ≥ 0.3, and G_R ≥ 0.3)—they should be employed as structural fills only if improved, for example, by preloading, because of their tendancy to exhibit high crushability. Otherwise, they can be safely used as general fills.
• C mixtures (R_B < 0.3 and G_R < 0.3)—likewise, GC mixtures should be employed as structural fills only if improved due to the high crushability of the glass. Otherwise, they can be safely used as general fills.
• S mixtures (R_B ≥ 0.6)—not only are they highly compressible, but from an environmental viewpoint are of concern due to the high concentration of metal leachate from the rubber inclusions and the associated risk of contamination to groundwater and soil [40,41]. While they could be potentially suitable as general fills (from a geotechnical viewpoint), they must be pretreated to reduce their environmental impact before being used as filling materials. Alternatively, they must be enclosed by impermeable geosynthetic membranes to prevent contamination of the surrounding environment [40,41].

6. Conclusions

This paper reported on the preliminary results of a feasibility study aimed at evaluating compaction, one-dimensional compressibility, and shear strength of recycled crushed green glass bottles and recycled granulated tire rubber mixed with gravel. Dry specimens of selected gravel–glass–rubber mixtures (GGRM), prepared at 0.4 and 0.55 (i.e., 40% and 55%) rubber content by volume (R_B) with varying glass content by volume (G_L = 0.15–0.6), were tested in the laboratory. The following main conclusions can be drawn from the study:

• Compaction—the dry density of compacted GGRM decreases linearly with increasing G_L and R_B. This is mainly due to the different specific gravity of the gravel (G_S = 2.66), green glass (G_S = 2.50), and rubber inclusions (G_S = 1.15; lightweight). Moreover, an increase in glass breakage/crushing with increasing G_L was observed, with the maximum amount of breakage (2.24%) obtained for R_B = 0.4 (G_L = 0.6). This is associated with the higher amount of glass in the mixtures. The glass breakage, however, seems to decrease with increasing R_B in the mixture; this is because soft rubber inclusions in the GGRM act as a cushion that absorbs part of the impact compaction energy delivered to the specimen.
• One-dimensional compression—the volumetric strain of the studied GGRM significantly increased with G_L due mainly to the glass crushability. Nevertheless, if the mixtures are subjected to unloading and reloading, their compressibility is drastically reduced, and their volumetric response becomes essentially elastic. This implies that during the unloading–reloading phase, the glass crushing and gravel grain rearrangement (mainly responsible for the plastic deformation) are minimized, and that the observed elastic response is due primarily to the rubber rebound.
• Shear strength—the shear strength of GGRM evaluated by direct shear tests is comparable to that of gravel–rubber mixtures (GRM) without glass in terms of overall mechanical response, peak shear strength, and friction angle. Specifically, under normal stress up to 100 kPa, the peak friction angle is found to be between 30° and 45°.
• Behavioral zones—seven distinct behavioral zones can be defined for GGRM due to the simultaneous presence of soft rubber-like (S), stiff gravel-like (G), and crushable glass-like (C) materials. The investigated compacted GGRM can be considered intermediate materials; their behavior gradually transitions from stiff–soft (GS) to stiff–soft–crushable (GSC) to stiff–crushable (GC) as glass is added in the mixtures as a replacement for gravel.
• Practical applications—overall, GGRM possesses adequate strength (i.e., a peak friction angle of 30° or greater) to be used as structural fill in eco-friendly geotechnical projects, such as lightweight backfill materials, shallow underground layers for mitigation of liquefaction phenomena, and geotechnical seismic isolation for low-to-medium-rise lightweight structures/infrastructure. While the volumetric compressibility of GGRM may be of concern in applications where high static loads are applied, this study demonstrates that it can be effectively enhanced (for example, by using preloading) and its impact minimized. Additionally, GGRM can be safely used as general fill material.

While this study only dealt with the response of GGRM under static loading conditions, the authors recognize that one of the critical components of material characterization is understanding their behavior under dynamic loading conditions. Although no experimental data are yet available for GGRM, investigations on GRM by the authors [60,61] and other researchers [62,63,64,65,66] may also provide useful information relevant to GGRM, such as small-strain stiffness and strain-dependent dynamic properties. Nevertheless, similar investigations are required to further characterize GGRM from a dynamic viewpoint.

Because the durability/aging of rubber may be a major concern, testing on GRM made with thermally aged rubber inclusions has been the focus of an ongoing long-term study carried out by the authors. The results of this investigation will be published elsewhere in due course and will serve as a reference for understanding the potential impact of aging on the response of GGRM with aged rubber inclusions.