An optical method for characterizing carbon content in ceramic pot filters

Household water treatment (HWT) is a common way for improving drinking water quality in terms of fecal pathogen disinfection in developing areas. Sobsey et al. (2008) and Hunter (2009) identified ceramic pot filters (CPFs) as one of the most effective available HWT methodologies. CMWG (2011) credits Fernando Mazariegos of the Instituto Centro Americano de Tecnología Industrial in Guatemala with developing the modern CPF design. That design consists of forming a mixture of native clay soil, a fugitive pore former (commonly sawdust), and water into a shape resembling a large flowerpot, firing the clay body in a ceramic kiln, and applying a silver compound solution to the final product. Salvinelli & Elmore (2015) reported that a typical CPF raw water volume is 9 L, and CMWG (2011) reported that CPF factories used a range of batch flowrate values of 1–3 L/h as the principal quality control criterion. A major advantage of CPFs is that the raw materials (with the exception of the silver solution) are typically available locally, and the production process can effectively be implemented using local labor.

CPF studies can be generally classified into the following categories based on the type of filter:

• production CPFs manufactured in developing countries (van der Laan et al. 2014; Salvinelli et al. 2016; and others);

• experimental CPFs fabricated in laboratories using either native clay soils or engineered clay (Plappally et al. 2011; Yakub et al. 2013; and others);

• disk or other non-pot shaped filters fabricated in the laboratory (Oyandel-Craver & Smith 2008; Ren & Smith 2013; and others).

It is common for researchers to characterize the porous nature of the completed ceramic body (Yakub et al. 2013; Soppe et al. 2015, and others) using the explicit or implicit assumption that the kiln firing sequence effectively combusts all of the fugitive pore former so that the final filter consists only of the vitrified ceramic body and the applied silver solution. Therefore, the firing sequences (temperatures and times) of factory kilns are not widely reported in the literature. Certainly, the effective construction and operation of a CPF kiln is readily recognized by manufacturers, and CMWG (2011) provides a relatively detailed description of kiln construction and firing variables. That document discusses means of kiln fire control including visual observation of the filters during firing, the use of pyrometric cones, the use of pyrometers and thermocouples, and the development of a written firing log. CMWG (2011) also describes draw trials as a means for confirming that the pore former has been completely removed from the production CPFs without requiring the destructive testing of the production filters. The interior or core of an incompletely fired CPF will include a layer colored black by carbonized pore former. A draw trial includes the insertion of incompletely fired CPFs pieces in a kiln, and subsequent manipulation of the firing parameters until fracturing of a trial shows that all of the carbon (by visual inspection) has been removed. By inference, CMWG (2011) states that proper firing of a CPF includes complete carbon burnout and fugitive pore former removal.

The authors observed the CPF manufacturing process at a factory near Antigua, Guatemala. The sawdust is past through a wire mesh to eliminate particles larger than a No. 8 sieve and is then mixed with clay from Rabinal, Guatemala at an approximate weight ratio of 6:1 clay-to-sawdust and wetted to create a working body. When it appears homogenous via visual inspection, the working body passes through an extruder to finalize mixing. The working body is then formed into a pot shape with a mechanical press. The shaped CPFs are then set to dry completely on shelving in the open-air factory for 1–3 weeks before being fired in a kiln. The filters are soaked in water overnight to saturate prior to a quality control flowrate test. A 1-hour falling head test is conducted in duplicate to ensure that the flowrates of the filters are between 1 and 2 L/h. Filters that pass the test are painted inside and out with a silver solution prior to distribution to consumers.

The use of activated carbon as a HWT to improve esthetic water quality has been marketed in the USA for several years by several manufacturers. A comparison of commercially available HWT systems by Al-Haddad et al. (2015) showed that an unspecified carbon-based filter (available in the USA) was reasonably effective at improving water quality including reducing free chlorine concentrations and providing bacterial disinfection. This study, along with others, may be the basis for the Guatemala CPF manufacturer's marketing statements that the presence of a carbon core in CPFs is a benefit to purchasers of their product even though CMWG (2011) explicitly describes a methodology for eliminating the core.

There are two primary questions associated with the presence of carbon cores in production CPFs such as the one shown in Figure 1: does a carbon core (incomplete firing) impact the efficacy of the CPF, and does the core provide a supplemental treatment benefit? A first step in addressing these questions is the identification of a means to characterize the carbon core at CPF production locations where Western laboratory services may not be available, may be prohibitively expensive, or may be logistically restrictive in terms of sample preservation and shipment. This paper describes an optical methodology that may have the practical potential for characterizing the carbon content of a production CPF without laboratory analyses. The paper also demonstrates the use of the optical methodology in a preliminary study of the impact of the carbon content on filter disinfection efficacy and flowrate.

This study used disk-shaped ceramic filters fabricated at the Missouri University of Science and Technology. The composition of the filters was selected to be both representative of the clay soil used by the Guatemala factory and to have a color that would provide a reasonable color contrast with any carbon core.

The clay body used in this study used the formula of the surrogate developed by Hubbel et al. (2015) to resemble the clay from Rabinal, Guatemala used in the CPF factory. The lower levels of iron in the surrogate clay gave it a lighter color relative to the native Rabinal clay soil and was beneficial for this study because it provided a contrasting background relative to the dark carbon core. Yellow pine sawdust was sieved through a No. 8 sieve and combined in a 6:1 weight ratio with the clay, then mixed with deionized water until homogenous. Experimental filter disks were formed with a 10 cm diameter manual press set to a thickness of 2 cm which was the typical thickness of the bottom of the filters produced in Guatemala. After pressing and before firing, the filters were allowed to dry for at least a week and spent at least 8 hours drying in a soils oven set at 100 °C. The filters in this study were not applied with colloidal silver in an effort to simplify the focus of the study because CPFs have been found effective even without silver present (Oyandel-Craver & Smith 2008; Clark & Elmore 2011).

The filters were fired in Cone Art Kilns, Inc. Model No. BX2327D electric kiln in three separate firings using identical schedules (temperatures and times) but with different atmospheres in order to develop multiple carbon core thicknesses. The firing schedule slowly increased the kiln temperature to 993 °C over the course of 14 hours before allowing it to cool back to room temperature. Ten filters were fired in each firing. For the mid-range atmosphere firing, oxygen was limited locally by placing the filters in a cordierite sagger box with a partial seal, and placing other combustible material (additional sawdust) in the box to partially scavenge oxygen that entered the box. The resulting atmosphere limited oxygen to result in incomplete burn out and leave some carbon remnants in the completed filter. For the reducing atmosphere firing, the process was repeated with additional combustible material in the sagger box and a thicker seal. The filters in this firing experienced less burnout than the first firing with a thicker carbon core. The oxidizing atmosphere firing was performed in air (without a sagger box), causing all of the pore-forming sawdust to fully combust.

The carbon content was analyzed both optically and by laboratory analyses. Laboratory samples were collected by powdering approximately 5 g of the inner perimeter of the 10 cm disks after the previously described 5 cm disks had been removed. The powder was then sent to the University of Missouri Extension Soil Testing Laboratory in Columbia, Missouri for total carbon analysis by combustion using an Elementar Vario MAX Carbon/Nitrogen Analyzer (Pella 1990; Nelson & Sommers 1996). The total carbon by combustion analysis is similar to methods used by Eliche-Quesada et al. (2011, 2012). Optical testing was performed by taking three images along the circumference of the 5 cm disks with both a Leica S8APO stereo microscope and a HiROX KH-8700 digital microscope. Two different instruments were used in order to validate the optical methodology. The Leica exposure was set at 12.6 ms with 2.1× gain, 100% saturation, and 1.00 gamma. The HiROX exposure was 1/250 s with 3 dB gain and white balance set at red 184 and blue 139. Each image was cropped to include only the filter material and then converted to greyscale. MATLAB 8.4 code was written to import the images as an array with a value ranging from 0 for black to 255 for white for each image pixel, and then the code returned the median value of the array as the tone value for the image. A paper standard was used to compare the performances of the two instruments. The standard had six printed blocks of color: white, white darker 5%, white darker 15%, white darker 50%, black lighter 25%, and black. An image was taken of each color block on each imaging device to identify a representative tone value for each block.

Seven samples from production filters made at the factory in Guatemala were also analyzed for carbon content. The cross sections of the samples contained a dark carbon core similar to that shown in Figure 1. The presence of the core indicated that complete combustion of the pore former had not occurred even though there was no visual evidence of carbon core on the surface faces of the samples. Three images were taken from the cross-sectional surface of each CPF sample to be representative of a filter that had experienced full burnout. The cross section of each sample was imaged seven times to characterize the carbon core, and the surface face of each sample was imaged three times to provide a background greyscale value for the ceramic body. The images were taken on both the Leica stereoscope and the HiROX microscope, and were processed in the same way as the experimental filter images to obtain a representative tone for each CPF sample.

Apparent porosity was characterized using a modification of the Archimedes method (ASTM C830-00). Two experimental filters, one with complete burnout and one with incomplete burnout with apparent high carbon content were tested. Instead of using a vacuum and pressure chamber to saturate the samples, each sample was left to soak submerged in water for at least 18 hours to achieve saturation. This procedure is similar to the procedure followed by the CPF factory in Guatemala to saturate the production filters before flowrate testing.

To test the experimental filters, a simulated contaminated surfaced water was produced using influent from the municipal wastewater treatment plant (WWTP) in Rolla, Missouri. The WWTP influent was mixed with tap water to create the testing influent water (‘challenge’ water) for the filters, which was passed through the filters for 4 weeks. The experiment length of 4 weeks was chosen to allow sufficient time for potential microbial growth using fugitive pore former as a carbon source. The challenge water coliform concentrations were on the order of 10² colony forming units per 100 mL (cfu/100 mL) for the first week, on the order of 10³ cfu/100 mL for the second and third weeks, and on the order of 10⁴ cfu/100 mL for the fourth week. These levels of contamination have been observed in surface water sources in Guatemala (Elmore et al. 2005). Challenge water total coliform and Escherichia coli levels were measured initially with Colilert Quantitray 2000^® but after the first week were measured with Coliscan Easygel^® in triplicate as the coliform concentrations grew too high to be quantified with the Quantitrays (which read up to 2,419.6 cfu/100 mL). Effluent water was tested for total coliforms and Escherichia coli with Colilert Quantitray 2000^® each time the concentration of coliforms in the challenge water increased to check for bacterial breakthrough.

After 4 weeks of testing, more highly contaminated challenge water (on the order of 10⁵ cfu/100 mL) was filtered to force failure so that the logarithmic reduction value (LRV) could be calculated. LRV is a measure commonly used to measure the bacteria removal efficacy of point of use (POU) water treatment systems (van Halem et al. 2007; Sobsey et al. 2008; Clark & Elmore 2011). The influent water was the same concentration for all filters and was tested for total coliforms and E. coli using Coliscan Easygel^® in triplicate. The effluent from each filter was tested for total coliforms and E. coli with Colilert Quantitray 2000^® to quantify the LRV.

Summary statistics for the experimental filter physical carbon content analysis, optical analyses, flowrate tests, and efficacy results in terms of log reduction values are given in Table 1.

The mean carbon content increased as the reducing atmosphere in the kiln increased. However, the variability of the carbon content as measured by the coefficient of variation (CV) did not directly correlate to changing atmosphere conditions, and the mid-range carbon content samples was the only data set with a CV greater than 1. The mid-range firing group was a highly varied group with a CV of 1.17, the only group with a CV greater than 1. Three of the filters in the mid-range group had lower carbon contents, and these three filters were closer to the kiln vent, which was partially opened to moderate the firing atmosphere. The individual filter positions in the kiln were not meaningful in terms of carbon content variability for the other two sets of filters because the vent was either fully open (oxidizing) or fully closed (reducing).

The porosity was measured for one oxidizing filter and one reducing filter. The apparent oxidizing filter porosity was 44.3%, and the reducing filter porosity was relatively similar at 43.8%.

A logarithmic relationship could be identified between the median tones generated using the optical technique and the percent carbon in experimental and production filters. The results were similar for both optical instruments used in the study, but the fit of the equations was relatively poor due to variability in the carbon content of the mid-range atmosphere samples which was largely attributed to the position of the filters in the kiln.

The variability of the mid-range atmosphere samples was also reflected in the LRV and flowrate analyses. However, the data from the initial study did not identify mathematical relationships between carbon content and disinfection efficacy or filter flowrate. More specifically, the carbon contents of the oxidizing atmosphere filter samples were relatively similar while the corresponding LRVs and flowrates had relatively larger ranges of values. The same observations apply to the reducing atmosphere filters. While it is tempting to conclude that flowrate and LRV are insensitive to the carbon content of a filter, the sample sizes used in the preliminary study are too small for definitive characterizations.

This initial study suggests that production CPF performance may be more tightly controlled with regard to LRV and flowrate if the pore former is fully combusted (no carbon core) via manipulation of the firing sequence to create an oxidizing atmosphere in the kiln. This implies that manufacturers may be able to produce more consistently performing CPFs if they maximize pore former burnout.

Finally, it is reasonable to assume that production facilities will not be able to control the firing atmospheres to the degree that those atmospheres can be controlled in the laboratory. Given the variability of carbon content and filter performance associated with position within the kiln for the mid-range atmosphere included in this study, it is reasonable to recommend additional studies regarding kiln packing.

The authors thank the management and staff of the CPF factory near Antigua, Guatemala, who provided the opportunity for observation of their production process.