Morphology, composition and performance of a ceramic filter for household water treatment in Indonesia

About one-eighth of the world's population does not have access to safe drinking water ((World Health organization) WHO/UN-Water 2010). Especially in developing countries, people suffer from waterborne diseases (WHO 2008). Point-of-use (POU) water treatment is often the only practical way for the local population to get safe drinking water.

One such developing region is the karst area of Gunung Sewu in Southern Java (Haryono & Day 2004). The poor water-retention capacity of the karst setting leads to acute water shortages, especially in the dry season (Nestmann et al. 2009). Therefore, the population depends on water from the karst aquifer, which was made available in a German–Indonesian joint project. However, this water is highly contaminated with fecal and other hygienically relevant bacteria, as the karst setting has a very low retention capacity for pollutants (Drew & Hötzl 1999, Goldscheider 2005). Thus, the water needs to be treated to prevent illness. Appropriate and sustainable water treatment concepts were developed within an Integrated Water Resources Management project funded by the German Federal Ministry of Education and Research. Considering the economic and technical circumstances of the region, low-cost solutions with simple technical requirements were chosen.

On the basis of previous work, water treatment was divided into three parts (Matthies et al. 2014), including a final step at household level. One of the most effective and appropriate technologies for drinking water treatment in poor regions is the use of ceramic water filters, like those promoted by the NGO Potters for Peace and approved by WHO (WHO/UNICEF 2012). These are manufactured and used in many developing countries (e.g. Cambodia, Indonesia, Ghana and Nicaragua), and exhibit good retention capacities for bacteria. A known problem with some POU ceramic filters is heavy metal leaching (Tun 2009 and Van Halem 2006).

In Indonesia, this kind of POU ceramic filter is produced by Pelita Indonesia (Pelita), Bandung, but promoted mainly in Western Java. According to Pelita, the filters are made of local clay, mixed with sawdust or rice husks depending on availability and price, and fired at 832 °C. The mixture ratio is adjusted when the clay source or pore-forming agents are changed. Fired filters are checked for flow rate and coated with silver nitrate, and one filter per batch is checked for removal of coliform bacteria and E. coli (Pelita 2012). However, no data are available on the chemical composition of the filters produced in Bandung and almost none about their effects on water quality. To determine the extent of heavy metal leaching, microbiological performance and other effects of the filter on the treated water, several filters were bought from Pelita for field studies in Indonesia and an extensive laboratory study in Germany.

To determine the structure of the filters, several filter fragments that were not coated with silver nitrate were analyzed. Scanning electron micrographs were obtained of all sides of one fragment with an environmental scanning electron microscope (Philips XL-30 ESEM-FEG, FEI Company, Hillsboro, USA) using a gaseous secondary (GSE) detector system. The acceleration voltage was 20 kV and chamber pressure about 120 Pa (0.9 Torr). The pore size of the filter is very important for its bacteria removal capacity, so a pore size analysis of two fragments was conducted at the (Forschungsinstitut für Anorganische Werkstoffe – Glas/Keramik – GmbH) Research Institute of Glass and Ceramics using mercury porosimetry, to get an idea of the pore size distribution.

Analysis for chemical composition was carried out using bulk powder from an uncoated fragment. X-ray diffraction analysis (XRD) gave information about chemical composition, especially the qualitative phase analysis of the ceramic. The diffractometer (Bruker/Siemens D5000, Bruker-AXS, Karlsruhe, Germany) uses CuKα radiation and is equipped with a graphite secondary monochromator. The elemental composition of the filter was analyzed using energy dispersive X-ray fluorescence ED XRF (Epsilon 5, PANalytical, EA Almelo, Netherlands). Bulk powder samples were sealed with a 6 μm Mylar film in spectro cups. A tungsten X-ray tube was used as the radiation source, with a Ge-detector for detection and quantification.

To analyze the filter's chemical performance, a series of experiments with tap water (TW) and natural river water (RW) were conducted. These were done with the original silver nitrate coated filter obtained from Pelita. First, 3 L of TW was applied to the filter and the filtrate (F1) was kept for further analysis. This experiment was repeated with 3 L for the second filtrate (F2) and 5 times with 5.5 L each time, for F3–F7. This series was performed mainly to determine any leaching effects from the filter, and the experiment was repeated subsequently with 6 L of RW (River Rhine at km 362.07, Germany, 8 January 2013) to analyze any removal capacity of the filter. For all experiments, several parameters were determined directly from the raw water and the filtrate: pH (pH/EC/TDS meter, HANNA Instruments Deutschland GmbH, Kehl am Rhein, Germany), electrical conductivity (WTW LF 330, Weilheim, Germany), nitrate and hardness (both: Hach Lange, Düsseldorf, Germany). To determine any metal leaching, the concentrations of metal ions were analyzed in both raw water and filtrate. The focus was on metals that were detected previously in the filter (see chemical composition, this work): As, Ba, Cd, Cr, Mn, Pb and Zn, as well as Ag because of the silver coating. Therefore, 100 mL of each filtrate was applied to a cation exchanger (Cation exchanger I, Merck KGaA, Darmstadt, Germany). Ions were then eluted with 5 mL 4 M nitric acid. The concentrations of metal ions were determined using inductively coupled plasma optical emission spectrometry (Optima 8,300 DV ICP-OES Spectrometer, PerkinElmer, Rodgau, Germany). Analysis for arsenic was executed by the Chemical Institute Pforzheim GmbH (CIP) according to the German drinking water regulation DIN EN ISO 11885, E22 (DVGW, 2001).

For the experiments with natural RW, turbidity (WTW Turb 355 IR, Weilheim, Germany) and total organic carbon (TOC) concentrations (Multi N/C^® 2000, Analytik Jena AG, Jena, Germany) were also determined. Turbidity was determined directly for both raw and filtered water. For the TOC analysis, the water was first passed through a 0.45 μm filter (Puradisc 25 TF; Whatman, Lawrence, USA) to remove bigger particles that would disturb the measurement.

To evaluate the suitability of the ceramic filter for use at the household level, the flow rate of filtrate was determined during the experiments.

In addition, water samples were taken from field experiments in Indonesia and analyzed at the CIP. Raw water, the filtrate of an old filter that had been in daily use for 2 years and the filtrate of a new filter were analyzed.

Liquid broth, agar and other chemicals were purchased from Merck KGaA (Darmstadt, Germany), Carl Roth GmbH & Co. KG (Karlsruhe, Germany), Applied Biosystems (Darmstadt, Germany), Sigma-Aldrich Chemie GmbH (Taufkirchen, Germany) or Oxoid Deutschland GmbH (Wesel, Germany).

To detect the effectiveness of bacteria removal, TW spiked with E. coli (DSM No. 1103, DSMZ GmbH, Braunschweig, Germany), Pseudomonas aeruginosa (ATCC 27853, LGC Standards GmbH, Wesel, Germany) (both Gram-negative) and Enterococcus faecium (DSM 20477, DSMZ GmbH) (Gram-positive) was filtered. Bacteria were cultivated in 30 mL liquid broth (12.5 g/L LB Broth) at 37 °C for 17 h in a shaking incubator. The cultures were then mixed with 7 L of TW and 5 L were applied to the filter. A culture technique with species-selective agar was used to examine the raw (RAW) and filtered (F) waters. P. aeruginosa was detected using Cetrimide agar, E. coli using ChromoCult^® Coliform Agar and Ent. faecium using Chromocult^® Enterococci Agar. Plates were incubated at 37 °C for 24 hours.

As recommended in the literature (WHO 2011c), the coliphages ϕX174 (DSM No. 4497) and MS2 (DSM No. 13767) were used as surrogates for viruses in the retention experiments, so 6 L of TW was spiked with a ϕX174 or MS2 phage suspension, to attain a phage concentration of about 10²–10³ PFU/L and 5.5 L of this was then applied to the filter. The remaining raw water (0.5 L) and 1 L of filtrate were concentrated and filtered through a mixed cellulose ester membrane at 550 mbar. Phages were then eluted and serial dilutions were made of the eluates for the plaque assay. The plaque assay was performed with the appropriate host E. coli cultures (DSM No. 13127 for ϕX174 and DSM No. 5695 for MS2). Plates were incubated at 37 °C for 24 hours. The TW was also mixed with phages and 0.08 g/L loess (Heilerde 1, Luvos^®, Germany) to determine any possible effect of turbidity on phage removal efficiency.

Experiments were conducted in duplicate, and analysis of the raw water and filtrates was again done in duplicate.

The filter is shown in Figure 1. It is reddish and frustum shaped, and its capacity was about 7.8 L.

ESEM pictures showed that the filter surface (not presented here) was relatively smooth. The broken edge of the filter (Figure 1) showed characteristic ceramic pores and crystal structures, as well as tiny, tube-like structures produced when pore-forming agents like sawdust or rice husks burned out completely during firing.

Mercury porosimetry was used to determine pore size distribution. The first filter fragment had a total porosity of 49.55% while, in the second, it was 49.24%. The fragments showed a substantial group of peaks between 1 and 40 μm with a smaller group from 0.006 to 0.04 μm (Figure 2).

The porosimetry showed that the pores are too big to retain bacteria by size exclusion. However, several mechanisms such as depth straining (Hutten 2007), particle bridging (Cheremisinoff 2002), inertial impaction, interception, diffusion and electrostatic attraction (Hutten 2007) play important roles in particle capture in depth filtration. The disinfection mechanisms of the silver coating also contribute to bacteria reduction.

A qualitative X-ray analysis of the filter showed that the existing mineral phases were quartz (SiO₂), hematite (Fe₂O₃), tridymite (SiO₂) and mullite (Al₆Si₂O₁₃), which was also confirmed by ED XRF. Moreover, ED XRF showed that some metals were present in relatively high concentrations and could, potentially, have a negative effect on human health if leached out in high concentrations (Table 1). They were, particularly, As, Ba, Cd, Cr, Mn, Pb and Zn.

As an uncoated filter fragment was used for this analysis, the Ag concentration was quite low, it will be higher in the coated filter.

The ceramic contained a relatively high amount of Fe₂O₃, suggesting an iron-rich clay, which produces the filter's reddish color. The arsenic in the ceramic probably also came from the clay, although rice husks used as pore-forming agents might play a role. It is known that rice grown on arsenic-rich soil and/or with contaminated water might also contain relatively high concentrations (Bhattacharya et al. 2009). Further analysis of the source materials would be necessary as a basis for recommendations for production of filters containing less arsenic.

To analyze the filter's chemical performance, several parameters were measured in raw and filtered water. Initially, the effect of the filter on laboratory TW was analyzed – see Table 2. The pH of the filtered water was slightly higher (pH 8) than the raw water (pH 7). No significant changes were noted in electrical conductivity (EC), hardness or nitrite concentration. For nitrate, there was a high of 12 mg/L in the first filtrate F1, but subsequent elutions showed no concentrations significantly higher than those in the raw water. Arsenic, manganese and silver were all leached from the filter in significant concentrations. Arsenic was present at about 0.09 mg/L initially, although concentrations decreased in subsequent batches of filtrate. In F7, the concentration was down to 0.018 mg-As/L. Manganese was leaching from the filter with a peak of 0.2 mg/L in the first filtrate, F1, and rapidly decreasing concentrations after that. Silver concentrations in the filtrate were about 0.01–0.03 mg/L, and were up to 26 times higher than those in the raw water.

The second set of experiments was carried out with natural RW and the results resembled those from TW. However, there was a significant reduction in turbidity from 2.3 NTU (raw) to 0.1 NTU (filtrate) – see the first part of Table 3 (columns headed RW and RF).

The second part of Table 3 – columns IW, IF-od and IF-new – shows the results of a field study in Indonesia. While 7.5 L were applied to the filters, the old filter, in daily use for about 2 years, had a flow rate of about 0.1 L/h. The new filter's initial flow rate was about 1.2 L/h. As in the laboratory experiments, turbidity, TOC, arsenic, manganese, nitrate and nitrite were washed from the new filter in the first filtrate. It is remarkable that the old filter reduced the EC of the filtered water from 504 to 270 μS/cm, while turbidity was reduced to 0.5 NTU and TOC to 0.2 mg/L. However, as the samples taken in Indonesia could not be stabilized for transport, the results give an idea on the water properties but cannot be regarded as 100% reliable. Thus, the reported reduction in EC might be influenced by adsorption mechanisms during transport, although both samples (raw and filtrate) were transported to Germany in the same way. Increased reduction in EC and hardness by the old filter might be due to changes in the pore surface of the filter that led to an increased reduction of ions. This was also confirmed by calcite values which were about 98 mg/L in the raw water and 54 mg/L in the filtrate of the old filter (new filter 65 mg/L) (data not shown).

The analysis for physico-chemical performance showed that pH, EC and hardness were consistent with guideline values. The same applied to nitrite (NO₂), whose concentration never exceeded guideline values. The initial leaching of nitrate (NO₃) is probably due to the silver nitrate coating, although it did not exceed the guideline value of 50 mg- NO₃/L (WHO 2011a). It was clear that arsenic, manganese and silver were leached from the filter. There are no guideline values for silver in the WHO drinking water regulations. However, silver concentrations of 0.05 mg/L or higher can be observed in drinking water disinfected with silver (WHO 2011a). Thus, silver concentrations of 0.02 mg/L are acceptable in the filtrate. The health-based recommended maximum value for manganese of 0.4 mg/L (WHO 2011a) was never reached. The only hygienically critical value in the study was that of arsenic in the filtered water, which was above WHO's threshold of 0.01 mg/L (WHO 2011a). However, the concentration of arsenic in the filtrate decreased after several liters of water had been filtered. In F7 (after 33.5 L), the value was down to 0.018 mg-As/L. The WHO guideline – 0.01 mg-As/L–is provisional, however, and WHO state that ‘… every effort should be made to keep concentrations as low as reasonably possible’ (WHO 2011b). In summary, a filter containing less arsenic would be desirable.

Another possible approach is preventing arsenic leaching from the filter by higher temperature firing or using fluxes like illites or feldspar, or a combination of both, to bind the arsenic into the glass phase. Further investigations are needed, particularly for those cases where the concentrations of leached arsenic or other metals are high. However, in case of the filter analyzed in this study, an intensive flushing of the filter before use reduces the arsenic concentration in the filtered water effectively.

During the TW experiments, the initial flow rate dropped from 1.82 (F3) to 1.2 L/h (F5), while 5.5 L were fed into the filter for each measurement (Table 2). For F2, only 3 L was applied, and the flow rate was less. The flow rate for natural RW was about 0.75 L/h when 6 L was applied, but it was found that the flow rate increased again when less turbid water was applied.

After 100 L of TW had been filtered the flow rate was about 1.4 L/h but it dropped to 1 L/h after 200 L were filtered. Following the equations of Schweitzer et al. (2013) describing the flow rates for the bottoms and sides of ceramic filters, and based on Darcy's equation, the hydraulic conductivities (k) of the filters in this study were between 1.5 and 4.4 × 10⁻⁷ m/s. This is higher than those observed by Van Halem (1.3–1.37 × 10⁻⁷ m/s) and Ebenezer et al. (1.46 × 10⁻⁷ m/s), but consistent with those described by Oyanedel-Craver and Smith, who observed hydraulic conductivities of 1.15–5.01 × 10⁻⁷ m/s (Van Halem 2006; Oyanedel-Craver & Smith 2008; Ebenezer et al. 2014).

Field observations showed that after 1 year (330 days) of daily use in Indonesia and an estimated volume of about 8,320 L filtered water, the flow rate decreased to 0.8 L/h.

Pelita aims to achieve 1.5–2.5 L/h (Pelita 2014). The initial flow rate in our experiments complied with this but the rate quickly dropped to about 1.3 L/h. Nevertheless, the filter produces well over 0.475 L/h, which is the minimum needed to produce sufficient drinking water assuming 24-hour usage, for an average family size (4 persons) and consumption of 2 Liters/person/day (Lantagne 2001). Even after 1 year of daily use in Indonesia the flow rate was still sufficient to supply a family of four.

Bacteria reduction experiments showed that the filter can remove about 5 log of E. coli, 3.4–4.5 log of Ent. faecium and 3.4–5 log of P. aeruginosa. Bacteria counts in the raw water and filtered water are given in Table 4.

Phage reduction experiments showed that the log removal was about 0.5–0.6 for ϕX174 and MS2 (Table 5). Turbidity in the TW experiments was much less than 1 NTU, with higher turbidity (8.5 NTU) the log reduction for ϕX174 was increased to 0.9.

Log reductions for bacteria were consistent with values described in literature such as those observed with E. coli by Lantagne et al. (2010), which were about 3.1–6.1. For phages, the log removal was comparable to values reported by Lantagne, which were about 0.09–0.5 log steps for filters produced in Nicaragua (Lantagne 2001).

The WHO guidelines for evaluating household water treatment options describe a bacteria log reduction of 2 as protective and 4 as highly protective (WHO 2011c). Thus, on average, the filter can be described as highly protective against bacteria. For viruses, the guidelines describe a log reduction of 3 as protective and 5 as highly protective (WHO 2011c). With a log reduction below 1, the filter exhibited very low removal efficiency for viruses. In the literature, it is noted that metal oxides mixed with the clay before firing might lead to a better absorption capacity for viruses (Brown & Sobsey 2009). One of the suggested metal oxides was Fe₂O₃, which is one of the main components of the filter tested. However, in this study insufficient virus removal was achieved. The effect of metal oxides in ceramic filters on virus removal needs to be addressed in further studies.

Morphological analysis of the filter showed that the pores, with sizes up to 40 μm, are too big to retain bacteria by size exclusion. Thus, depth filtration retention mechanisms and the disinfection properties of the silver coating must play an important role in bacteria reduction.

The results showed that silver, arsenic and manganese were leached from the filter. However, silver leaching was at acceptable concentrations, manganese was washed out after a few liters and arsenic concentrations in the filtrates rapidly decreased, too. Though leaching of arsenic did not pose a health risk in the filters analyzed in this study, methods of inhibiting leaching, such as higher firing temperatures, will be addressed in future studies.

The reduction in bacterial numbers was 3–5 logs, while phages were reduced by 0.5–0.9 logs. Thus, according to the WHO guidelines, the filter was protective against bacteria while its removal efficiency for viruses was quite low. Possible methods of optimization for virus removal need to be addressed in further studies.

In conclusion, the filter examined offers a relatively easy way of reducing bacteria in drinking water at household level. After discarding the first liters of filtrate, the water did not contain heavy metals leached from the filter in health-relevant concentrations.

The authors would like to thank the CIP Chemical Institute Pforzheim GmbH and the Research Institute for Inorganic Materials–Glass/Ceramics–GmbH for the great collaboration and kind assistance. We would like to thank Elisabeth Eiche, Frank Friedrich, Frank Kirschhöfer, Utz Kramer and Annett Steudel for their support. We want to thank the German Federal Ministry of Education and Research (BMBF) for financing this project and Pelita Indonesia for their kind assistance.