Characterization and evaluation of recycled glass sand as water filtration media Open Access

Water is essential for many industrial, agricultural, and household uses, including heating and cooling, irrigation, and consumption, making access to clean water a necessity for sustainable living environments (Lin et al. 2022). Thus, ensuring the availability of clean water was identified as a critical sustainable development goal (SDG) in the United Nations publication Transforming our World: The 2030 Agenda for Sustainable Development (UN 2015). However, industrialization and population rise have led to a substantial escalation of pollutants in water (Ighalo et al. 2021; Lin et al. 2022). Consequently, water must undergo adequate treatment before being utilized for any purpose (Twort et al. 2000).

Traditional methods for drinking water treatment involve multiple processes that may include screening, coagulation, flocculation, sedimentation, filtration, and disinfection. Coagulation and flocculation are chemical and physical processes, respectively, utilized to aggregate contaminants prior to removal by sedimentation and/or filtration. Sedimentation is effective when aggregated particles settle due to gravity (Centers for Disease Control and Prevention 2024), but filtration is often required to remove suspended solids and dissolved contaminants. Filtration through granular media is the predominant approach to eliminating suspended solids, reducing the presence of microorganisms in water, and enhancing the effectiveness of the disinfection process (LeChevallier & Au 2004; Environmental Protection Agency 2020; Centers for Disease Control and Prevention 2024).

Silica stands out as the primary material employed in water filtration, yet other commonly utilized granular media include anthracite coal, garnet sand, and ilmenite (Letterman 1999; Soyer et al. 2010; Huifang et al. 2020; Hoko et al. 2024). In general, finer particles within the filter medium obstruct the passage of larger particles in the water. Effective filter performance depends on the selection of an appropriate filtration medium (Cescon & Jiang 2020). The efficacy of the filter medium is influenced by properties such as size distribution, density, shape, and porosity (Cescon & Jiang 2020). A variety of filter media types can be employed individually or in combination, such as dual or multi-media filters, depending on specific needs (Soyer et al. 2010).

While silica sand is the most used material for filtration, it is not always the most feasible option either due to local geology or economic considerations (Soyer et al. 2013). The use of recycled glass as a filter medium has been investigated as a replacement for natural silica sand due to its availability, a shortage of alternative markets for recycled glass, and growing interest in sustainability (Horan & Lowe 2007). Research has shown that recycled glass sand can be a viable alternative to silica sand (Elliot 2001; Gill et al. 2009). Recycled glass has an advantage over other alternatives as it is less expensive, environmentally friendly, and can be pulverized to create filtration media of different particle sizes (Soyer et al. 2013). Studies comparing recycled glass sand to silica sand in a variety of scenarios, including swimming pool water cleaning (Korkosz et al. 2012), as a tertiary medium in wastewater treatment (Horan & Lowe 2007) and lagoon water filtration (Salzmann et al. 2022), have shown that recycled glass sand can achieve comparable turbidity reduction and removal of total suspended solids. To the best of our knowledge, the effectiveness of recycled glass sand as a filtration medium for drinking water treatment has not been studied. According to a survey conducted in 2016, glass accounts for 5% of global solid municipal waste (Kaza et al. 2018). However, the recycling rate of glass varies by region. For example, the glass recycling rate in Europe is 79% (Harrison et al. 2020) whereas the glass recycling rate in the United States is only 31% (Environmental Protection Agency 2019). Because glass is not biodegradable, it occupies valuable space in landfills. Poor recycling of waste glass also causes higher dependency on natural resources, which is leading to the depletion of sand resources worldwide. Growing demand for landfill space and the depletion of natural resources are driving the exploration of alternative methods for reusing waste glass (Harrison et al. 2020). Apart from recycling old glass into new glass, there are many other applications for which recycled glass could replace silica sand, which would reduce landfill waste and alleviate sand shortages. Crushed waste glass from food and beverage containers is particularly suitable for recycling as it can be useful with minimal processing and it is 100% recyclable (da Silva et al. 2019; Ingrao et al. 2021). For example, it contains no heavy metals (present in leaded crystal), chemical residues (possibly present in glass sourced from chemical storage containers or glass tubes for neon lights), or harmful microbes (only benign microbes) that may require additional processing steps to ensure safety.

The objective of this study was to evaluate the use of recycled glass as a filter medium for water treatment in a gravity-driven flow system on a scale appropriate for household water filtration. We build upon prior work that has shown crushed recycled glass can be as effective as silica sand in a variety of water treatment scenarios (Horan & Lowe 2007; Korkosz et al. 2012; Salzmann et al. 2022) and that layered multi-media filters can improve water filtration over single-medium filters (Hoko et al. 2024). Specifically, we analyzed the performance of different recycled glass sand particle sizes in single-layer and multi-layer configurations for small-scale water filtration (from 1 to 10 L). In contrast to the existing multi-media filtration studies, in which materials differ in both particle size and chemistry (e.g., silica and anthracite (Hoko et al. 2024)), we varied only particle size, allowing us to analyze this variable without the presence of chemical differences as a confounding factor. Additionally, prior studies examining recycled glass sand as a filter medium for environmental water samples measured effectiveness by turbidity, total suspended solids, and/or chemical oxygen demand (Horan & Lowe 2007; Soyer et al. 2010; Korkosz et al. 2012; Salzmann et al. 2022); to the best of our knowledge, our study is the first to evaluate a combination of layered glass sand filtration with polyaluminum chloride (PAC) coagulation pretreatment for removing bacteria in water. In this manuscript, we describe findings on the physical characteristics and hydraulic properties of crushed recycled glass with different particle sizes and how these properties correlate with the effectiveness of single- and multi-layer filtration systems with respect to permeability, turbidity reduction, and microbial removal.

The recycled glass sand used in this study was provided by Glass Half Full, a New Orleans-based start-up company. Glass Half Full collects soda lime glass (primarily food and beverage containers) for recycling through their drop-off and collection programs. Glass Half Full does not accept leaded crystal. Crushed glass received from Glass Half Full was separated by size: 3.4–1.7 mm, 1.7–0.4 mm, and 0.4–0.01 mm. The 0.4–0.01 mm fraction was further sifted in the laboratory using stacked sieves to remove particles <0.1 mm. Throughout the manuscript, these fractions are referred to as G (gravel, 3.4–1.7 mm), CS (coarse sand, 1.7–0.4 mm), and FS (fine sand, 0.4–0.1 mm). The effective size and uniformity coefficient (UC) were also determined for each fraction. The <0.1 mm particles were used as the probe material in the particulate filtration experiments.

To mimic the turbidity due to particulates in water, 0.75 g of <0.1 mm glass sand particles were added to 1 L of deionized water. To ensure uniform turbidity, the mixture was stirred continuously using a magnetic stirrer at 900 rpm. The pre-filtration turbidity was approximately 75 NTU for all particulate filtration experiments. Three replicates of water with glass sand particles were filtered through each configuration. The pre-and post-filtration turbidities were measured using a turbidity meter (HF Scientific, M 100 Plus).

Water samples were collected from Lake Pontchartrain, an estuary located in southeastern Louisiana in the United States to serve as an environmental matrix. To evaluate the effectiveness of the recycled glass sand in removing bacteria in water, 1-L replicates of lake water were seeded with Escherichia coli (American Type Culture Collection, ATCC 25922) (Yechouron et al. 1991) in triplicate, in separate experimental batches for each treatment combination and filtration setup. E. coli was chosen as a common fecal indicator organism due to its association with the presence of other pathogenic microorganisms (Environmental Protection Agency 2021). Before spiking, the indigenous E. coli concentration in the lake water was measured multiple times and found to be 8.05 ± 5.34 colony-forming units CFU/100 mL, which is below the target range for removal testing. Therefore, all samples were spiked with E. coli at concentrations ranging from 10¹ to 10⁴CFU/100 mL to achieve environmental relevance and detectability.

Three replicates were tested for both the single-layer and two-layer configurations. For the three-layer configuration, we conducted as many replicates as possible, totaling eight replicates. Additionally, to mimic the water treatment process, samples filtered through the two-layer and three-layer configurations were pre-treated with 15 mg/L of polyaluminium chloride (PAC). The coagulation process included rapid mixing at 125 rpm for 5 min followed by 90 min of settling before filtration of the supernatant. The PAC concentration and settling time were chosen based on jar tests conducted using lake water in which PAC concentrations of 100, 50, 40, 30, and 15 mg/L were assessed by turbidity reduction following methods reported by Nti et al. (2021). According to these tests, 15 mg/L PAC was selected for subsequent experiments, achieving over 80% turbidity reduction. The coagulation pH was 7.4 ± 0.5, and the water temperature was 19.0 ± 0.6 °C. The pH of environmental water (Lake Pontchartrain) used in the experiments ranged from 6.6 to 8.72, with an average of 7.5 ± 0.6. Prior to each experiment, sterile ultrapure water was filtered through the filtration setups to ensure the absence of E. coli in the glass material.

E. coli concentration and water turbidity were tested before and after filtration, and before and after PAC treatment when applicable. For PAC treatment, water samples were collected from the supernatant after sedimentation. For glass filtration experiments, samples were collected directly from the tap of the filtration setup. The flowrate was determined by dividing the volume of water tested (mL) by the filtration time (s).

The concentration of E. coli was determined using the mTEC ChromoSelect Agar, following the manufacturer's instructions and Environmental Protection Agency (EPA) method 1103.1 (Environmental Protection Agency 2010). Briefly, 50 mL of sample (or phosphate-buffered saline (PBS) as a negative control) was filtered through a nitrocellulose membrane filter of 0.45 μm pore size. After filtration, the membrane was placed in the mTEC medium and incubated for 24 h at 44.5 °C. The colonies were counted, quantifying the bacteria as CFU/100 mL.

Water turbidity and pH were measured using the HACH 2100 Turbidimeter and Hanna EDGE Multiparameter meter, respectively.

Statistical analysis and graphical representations were created in R Studio version 4.4.0. Log₁₀ bacterial reductions were calculated by subtracting the log₁₀ concentration of E. coli before and after water treatments. The turbidity removal percentage was calculated by subtracting the initial turbidity from the final turbidity and dividing the result by the initial turbidity.

Because all glass sand particles originated from soda lime glass, the particle density is the same irrespective of particle size. However, the bulk density decreases as particle size decreases (Table 1). Thus, porosity increases with decreasing particle size. The particle size, effective size and uniformity coefficient (UC) are also shown in Table 1. Low uniformity constant values are preferred for sand filtration ideally less than 4 (Tao & Mancl 2008; Verma et al. 2017). The UC values for all glass sand filters were less than 4.

To evaluate the efficacy of recycled glass sand in reducing water turbidity, the <0.1 mm fraction of recycled glass sand particles was suspended in water for particulate filtration tests. The particles in this fraction were mostly in the range of 0.06–0.08 mm with <10% of particles detected in the 0.03–0.06 mm range via a particle size analyzer (Malvern Multisizer 3000). Understandably, the trend in turbidity reduction is opposite to permeability in relation to glass sand particle size. Specifically, the turbidity reduction using G (≈ 35%) was much lower than that achieved with CS (≈ 94%) or FS (≈ 97%) (Figure 2(a)). However, between CS and FS, the improvement in turbidity reduction was marginal compared with the reduction in permeability, which decreased by a factor of two.

In comparison with other work, our CS medium has the most comparable particle size and porosity to silica sand media studied by others. In these studies, water filtration using silica sand typically produced a turbidity reduction ∼60%, although results varied based on the specific filtration setup and starting turbidity (Soyer et al. 2010; Davies & Wheatley 2012; Hoko et al. 2024). The work by Soyer et al. (2010) provides one of the more insightful comparisons for our work because they examined both silica sand (porosity of 0.38) and crushed glass (porosity of 0.49) as a single-layer filtration media. In their work, water samples with an initial turbidity of 5 NTU were filtered through a 104 cm filter bed. A turbidity reduction of 40–70% (5 NTU to 1.5–3 NTU) was achieved for both filter media, but lower head loss was seen for the recycled glass medium. Although we achieved a greater turbidity reduction with a much shorter column (10 cm), we believe this result was mainly due to a higher starting turbidity (75 NTU). However, another possibility is that the use of different water sources (deionized water in our experiments, and environmental water in experiments by Soyer et al. (2010)) also contributed to differences in our turbidity reduction outcomes. Environmental water may contain different types of particulate matter, including clay, silt, or microorganisms, and differ in ionic strength or pH from deionized water, and these characteristics play a crucial role in turbidity reduction performance. These explanations are supported by our testing with environmental water samples (next section) in which the starting turbidity was 13–40 NTU and the reduction was ∼60%, similar to that reported by Soyer et al. (2010).

To assess the individual performance of each particle size, single-layer filtration experiments were conducted without PAC pretreatment, allowing for a direct comparison of removal efficiency across different media particle sizes. In this study, initial lake water turbidity ranged from 13 to 40.3 NTU, and the seeded E. coli concentrations ranged from 4.6 × 10¹CFU/100 mL to 3.1 × 10³CFU/100 mL.

Based on the single-layer study results, a two-layer filtration system was tested for improving turbidity reduction and E. coli removal from environmental water samples without compromising the flowrate and filter permeability. PAC treatment was included prior to the filtration process in these tests. The three replicates of water samples seeded with E. coli had an average initial concentration of 1.1 ± 0.1 × 10⁴CFU/100 mL. After PAC treatment, the average E. coli concentration was 2.4 ± 1.3 × 10³ CFU/100 mL, indicating an average LRV of 0.7 ± 0.3. Following filtration, the average E. coli concentration was further reduced to 4.7 ± 1.7 × 10² CFU/100 mL, indicating an average LRV of 1.5 ± 0.1. The combination of PAC treatment and the two-layer setup achieved excellent bacteria removal, with an average LRV of 1.5, indicating approximately 96% reduction. Turbidity reduction was also notable: PAC treatment achieved an average turbidity reduction of 75.2%, while the combined PAC treatment + filtration process resulted in a final turbidity reduction of 86.6%. To isolate the effect of PAC, the two-layer configuration was also tested without coagulation. In the absence of PAC, E. coli log removal decreased to 0.27 ± 0.15, and turbidity reduction dropped to 33.3 ± 22.8%. Statistical analysis confirmed that PAC significantly improved turbidity reduction (p = 0.032, t-test), though the difference in E. coli log removal was not statistically significant (p = 0.077, Mann–Whitney U test). The improvement in turbidity reduction and E. coli reduction with the addition of PAC is due to the agglomeration of contaminants and forming larger flocs that are more easily captured by filter media (Mohamed et al. 2020; Diharjo et al. 2022).

Water quality improvements were expected from PAC treatment, as PAC is an inorganic coagulant widely used in drinking water treatment (Salzmann et al. 2022). Briefly, PAC hydrolyzes in water producing a positively charged aluminum hydroxyl complex that facilitates the flocculation and settling of negatively charged colloid particles. PAC aids in the sterilization process by electrostatically pulling bacteria into these flocs (Tang et al. 2016; Salzmann et al. 2022). Analyzing the turbidity reduction and the E. coli log removal across the replicates, it is evident that PAC treatment alone achieved moderate to good removal of bacteria and turbidity reduction. However, filtration after PAC treatment further enhanced the removal of E. coli and reduced turbidity across all seven replicates tested (Figure 5, Table 3). Thus, adding the FS layer to the bottom of the filter enhanced both turbidity reduction and E. coli removal with only a small reduction in permeability compared to the two-layer setup (Figure 3). Additionally, a Wilcoxon signed-rank test indicated that E. coli log removal was significantly higher after PAC + filtration compared to PAC treatment alone (p = 0.022), confirming the added value of the three-layer glass filtration system in microbial removal. This improvement likely occurred because the flocs formed during coagulation were too small for the CS size to trap them effectively.

Historically, the combination of coagulants and granular filtration has proven to be a highly effective approach for water treatment (Bratby 2016). In our study, the integration of PAC treatment and crushed glass filtration significantly improved water quality, demonstrating its reliability and potential applications in water treatment. These findings align with previous studies that have reported successful turbidity reduction and/or microbial removal using similar treatment combinations (i.e., coagulation plus filtration through media containing silica sand or crushed glass (Evans et al. 2002; Soyer et al. 2010, 2013; Davies & Wheatley 2012; Sabiri et al. 2017). Additionally, all environmental water samples filtered using the three-layer setup in our study showed final turbidity levels below 5 NTU, which complies with the World Health Organization (WHO) recommendations for drinking water quality (World Health Organization 2022). Although disinfection would still be needed to ensure drinking water safety, removing particulate matter and reducing the concentration of microbial species prior to disinfection enhances the process (LeChevallier & Au 2004; World Health Organization 2022).

Clogging is a multifaceted process characterized by the reduction of porosity and permeability within a filter bed, which reduces hydraulic conductivity (Lunardi et al. 2022). It can be categorized into three main types: physical, chemical, and biological. Physical clogging occurs when suspended sediment is retained within the filter bed, impeding the flow of water. Chemical clogging, on the other hand, occurs when chemical substances precipitate or accumulate within the porous medium. Biological clogging primarily affects the superficial layers of filter beds, where solids are more likely to accumulate. This environment favors the growth of microorganisms such as protozoa, algae, fungi, and bacteria (Le Coustumer et al. 2012; Grace et al. 2016; Lunardi et al. 2022). In our experiments, the more dramatic decrease in flowrate for the environmental water samples compared to the particulate water samples was likely due to the added growth of microorganisms in the filter bed subjected to environmental water. Further study will incorporate filter backwash in the system to prevent clogging.

Understanding the filtration configuration is important because it affects water filtration performance (Cescon & Jiang 2020). For example, small flocs may bypass a coarse filtration layer, leading to poor contaminant removal, but large flocs can rapidly clog a fine filtration layer, reducing the system's overall efficiency (Cescon & Jiang 2020). In our study, the decision to use a multi-layer configuration was driven by the different size distributions of particulates, flocs formed upon PAC treatment, and un-flocculated E. coli in particulate and environmental water samples. The three-layer configuration allowed the coarser G (3.4–1.7 mm) layer to trap larger species, preventing premature clogging of the CS (1.7–0.4 mm) and FS (0.4–0.1 mm) layers, while an FS layer was needed to remove the finest contaminants. However, keeping the FS layer short (1 cm) was necessary to retain a reasonable flowrate of water through the column. As particulate filtration results show, reducing the FS layer from 10 cm (single-layer FS experiments) to 1 cm (three-layer experiments) increases the flowrate by a factor of 6.7 (from 0.0009 m/s or 1.6 L/h to 0.0061 m/s or 10.8 L/h) while maintaining the same turbidity reduction.

This research demonstrates that recycled glass sand can be used as a water filtration medium. Its use may be most appropriate for application in household slow sand filters (HSSF) or similar systems. While a direct experimental comparison with silica sand was not conducted in this study, other studies have shown that HSSFs using silica sand, bio sand and ceramic filters achieve a turbidity reduction of 75–90% (Murphy et al. 2010; Jenkins et al. 2011; Young-Rojanschi & Madramootoo 2014; Singer et al. 2017), which is comparable to the turbidity reduction achieved in this study. Specifically, the three-layer filtration system achieved a turbidity value of under 5 NTU, which aligns with the values recommended by the WHO (World Health Organization 2017).

This study is, to the best of our knowledge, the first to integrate PAC coagulation pretreatment with a three-layer crushed recycled glass filtration system and evaluate its effectiveness for turbidity and E. coli removal in a household-scale context. The use of PAC significantly improved turbidity reduction (p = 0.032) and enhanced microbial removal. Furthermore, a Wilcoxon signed-rank test confirmed that E. coli log removal was significantly greater after PAC + filtration compared to PAC treatment alone (p = 0.022), validating the added value of this integrated approach. However, we acknowledge certain limitations in our study. Some treatment groups had limited replicate numbers, and although statistical methods such as non-parametric tests were used to evaluate significance, larger sample sizes would improve the robustness of the conclusions. Additionally, pH was not varied as an independent experimental factor, which limits our ability to isolate its effects on filtration performance.

With the rising frequency and intensity of natural disasters, regaining timely access to clean water has become one of the primary concerns in disaster-hit areas, particularly for difficult-to-reach and low-income communities. Our research suggests that recycled glass sand filtration units could be a potential short-term water purification solution in these situations.

This work was supported by the National Science Foundation Convergence Accelerator Program through Award #2230769. We gratefully acknowledge Glass Half Full in New Orleans, Louisiana, for providing the crushed recycled glass used in this work. We thank Dr Vijay John for the helpful discussion during the study design.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.