Granular Activated Carbon Adsorption of Carcinogenic Volatile Organic Compounds at Low Influent Concentrations

Abstract

The effectiveness of granular activated carbon (GAC) for carcinogenic volatile organic compounds (cVOCs) has not been evaluated in the low- to sub- microgram per liter range. Rapid small scale column tests (RSSCTs) were employed to determine the GAC performance at empty bed contact times (EBCTs) of 7.5 and 15 minutes for 13 cVOCs at a target influent concentration of 5 μg/L in a typical groundwater matrix. Breakthrough was assessed for vinyl chloride, dichloromethane, 1,1-dichloroethane, 1,2-dichloroethane, 1,2-dichloropropane, carbon tetrachloride, 1,3-butadiene, 1,1,1,2-tetrachloroethane, 1,2,3-trichloropropane, trichloroethylene and tetrachloroethylene. The throughput to breakthrough was found to be linearly correlated to capacities calculated with single-solute equilibrium isotherm parameters. Modest decreases, 9 to 13% on average, in throughput to 50% and 75% breakthrough were found when the EBCT was increased from 7.5 to 15 minutes. The carbon use rate (CUR), when scaled to simulate full-scale adsorption, indicated that GAC would be a viable technology for seven of the VOCs evaluated, with a CUR threshold less than 0.2 lbs/1000 gal. It may be possible to use 1,1 DCA and 1,2 DCA as surrogates for assessing chemicals near the feasibility limit.

Introduction

About 105 million people in the United States, over one-third of the population, receive their drinking water from public systems using groundwater (Toccalino and Hopple 2010). Over 19% of the groundwater samples from 3,498 wells sampled in a USGS study contained volatile organic compounds (VOCs) at a concentration of 0.2 μg/L or greater (Zogorski et al. 2006). Of 98 aquifers investigated, 90 contained VOCs (Zogorski et al. 2006). Another study detected VOCs in 60% of 833 groundwater samples that are used as a source of drinking water (Rowe et al. 2007). The U.S. Environmental Protection Agency (USEPA) currently regulates eight carcinogenic VOCs (cVOCs) for potential health effects from long-term exposure in drinking water. The maximum contaminant level goal (MCLG) for these cVOCs is zero and the maximum contaminant level (MCL) ranges from 0.002 to 0.005 mg/L.

The USEPA identified several emerging cVOCs (e.g. 1,2,3-Trichloropropane and 1,1- Dichloroethane) in the third candidate contaminant list, which was published in 2009. In 2011, the USEPA introduced a drinking water strategy to enhance public health protection by going beyond the traditional framework that regulates one contaminant at a time. Specifically, one of the strategies is to accelerate advancement of drinking water protection by addressing contaminants as a group (USEPA 2011). Additionally, with an MCLG of zero for the regulated cVOCs, recent improvements in the analytical techniques provided an opportunity to assess if the MCLs of these compounds can potentially be lowered because MCLs need to be set as close to the MCLG as is feasible with the use of the best technology. These two drivers have led to interest in studying a group of cVOCs that included both regulated and the emerging cVOCs. To assess technical feasibility to treat cVOCs to potential lower quantitation levels, this study was conducted to determine whether the designated Best Available Technologies (BATs) will cost-effectively treat these compounds to potential new quantitation levels. The BATs for the currently regulated cVOCs are granular activated carbon (GAC) and packed tower aeration (PTA).

GAC adsorption can be a viable water treatment technology for the removal of a variety of organic compounds and has been identified by the USEPA as a BAT for the currently regulated cVOCs, except dichloromethane and vinyl chloride (USEPA 2009). However, BAT determinations were based on studies carried out at relatively high influent and effluent concentrations, in accordance with current MCL concentrations on the order of 0.005 mg/L (5 μg/L). Furthermore, the drinking water strategy contaminant group’s effort adds eight additional cVOCs for consideration (USEPA 2011b). Previous work has evaluated the use of GAC to remove VOCs (Jarvie et al., 2005; Hand et al. 1989, Summers et al., 1989); however, these studies mainly focused on surface waters or utilized influent concentrations that were significantly higher than the sub microgram per liter range. Therefore, a data gap exists for BAT determinations at potentially lower sub-microgram per liter concentrations and for a larger list of cVOCs than has been previously investigated.

For compounds of health concern, including those regulated, desorption from the GAC back into solution can pose a problem because target treatment objectives, including MCLs, can be exceeded. Desorption can occur by two main mechanisms; change in the target compound concentration gradient and displacement by a competing compound (Corwin and Summers 2011). Generally, desorption via changing concentration gradient is not an issue because ambient target compound concentrations in groundwaters are relatively constant. However, displaced desorption can be an issue when other competing solutes are present, including fractions of the background dissolved organic matter (DOM) and VOCs which often co-occur in groundwater (Toccalino and Hopple 2010).

DOM is a mixture of organic compounds with a wide range of adsorbability, extending from non-adsorbable to highly adsorbable in nature (Chowdhury et al. 2013). The mass transfer zone of the more weakly adsorbable fractions moves through the column quickly relative to most target contaminants. As these compounds penetrate deeper into the GAC bed they can “foul” the adsorption sites or block the pore, causing adsorption capacity reduction for the target compound. This often manifests itself with earlier normalized breakthrough at longer empty bed contact times (EBCT) (Corwin and Summers 2010; Jarvie et al., 2005). In addition, DOM’s more strongly adsorbing fraction has a slower mass transfer zone and displaces previous adsorbed target compounds with a lower adsorption affinity (Snoeyink and Summers 1999). Therefore, displaced desorption, particularly with respect to DOM and the less adsorbable cVOCs of interest, requires investigation.

The overall objectives of this research were to evaluate GAC for the removal of a group of cVOCs at lower concentrations in the sub-microgram per liter range including additional cVOCs not currently regulated. Additionally, objectives included the study of adsorptive competition on the GAC between the cVOCs and DOM, which could become highly relevant at the low cVOC concentrations of interest. To accomplish the research objectives, multiple rapid small scale column tests (RSSCTs) were employed to determine the GAC performance for 13 VOCs at two EBCTs. Furthermore, the impact of both the background DOM type and concentration was examined.

Materials and Methods

Water Sources

The majority of the experiments were carried out with water collected from the Greater Cincinnati Water Works Bolton Water Treatment Plant Well #6 (OH). The well was always operated for at least 24 hours prior to collecting water in several high density polyethylene (HDPE) drums. The drums were stored at 4°C until needed and used within three months of collection. An additional two groundwaters used for comparative purposes (CO I and CO II) were collected from two sites in Boulder County, Colorado after pumping at least 20 well volumes prior to collection. Water quality parameters are shown in Table 1.

Table 1 –

Source Water Quality

Water	pH (−)	TOC (mg/L)	UVA254 (cm−1)	SUVA (L/mg-m)
OH GW	7.5	1.0	0.0155	1.55
CO GW I	7.0	0.3	0.0044	1.47
CO GW II	7.8	1.5	0.0202	1.35

Chemicals and Influent cVOC Preparation

The 13 compounds investigated in this effort were vinyl chloride (VC), dichloromethane (DCM), 1,1,2,2-tetrachloroethane (1,1,2,2 TCA), 1,1-dichloroethane (1,1 DCA), 1,2-dichloroethane (1,2 DCA), 1,2-dichloropropane (1,2 DCP), carbon tetrachloride (CT), 1,3-butadiene, 1,1,1,2-tetrachloroethane (1,1,1,2 TCA), 1,2,3-trichloropropane (1,2,3 TCP), benzene, trichloroethylene (TCE) and tetrachloroethylene (PCE). Compounds were investigated individually in all experiments. Regulatory MCLs, analytical detection limits, Freundlich parameters, minimum equilibrium concentrations (K_F,1/n, and Min C_e, Speth and Miltner 1990), and the equilibrium solid-phase loading (q₅) expected with an influent cVOC concentration of 5 μg/L are shown in Table 2. The chemicals were purchased neat, in the highest purity available, analytical standard grade (≥99.9%)¹ for the 11 liquid chemicals. The two gases, VC and 1,3-butadiene were also purchased in the highest purity available, puriss. grade (≥99.5%)¹. Neat chemicals were mixed with distilled deionized water to make stock solutions in the range 100 μg/mL to 600 μg/mL and stored headspace-free in multiple crimped-top 1.8 mL glass vials in a 4°C refrigerator. Gaseous chemicals were bubbled through distilled deionized water for several minutes and then the water was stored the same way. To make cVOC impacted water for OH groundwater experiments, 20-L groundwater was pumped headspace-free into a custom designed perfluoroalkoxy (PFA) bag² supported across two magnetic stir plates. A single stock solution vial was used to make a 5 μg/L experimental influent solution via direct syringe injection into the bag, stirred overnight, and connected to the experiment the next day. Preparing cVOC-impacted water for CO I and II experiments was done using 22-L glass carboys filled to near capacity. Direct syringe injection of stock solution was used to create a 5 μg/L experimental solution. This solution was then mixed using a Teflon stir bar for approximately 20 – 30 minutes before being used in an experimental RSSCT. A volatile trap (with approximately 15 – 20 μg/L cVOC) was incorporated into the CO I and II set-up to limit volatilization of cVOC from the glass carboy as water levels decreased and headspace increased in the carboy.

Table 2 -.

cVOCs and corresponding MCLs, analytical detection limits, Freundlich K and 1/n values, minimum equilibrium concentrations and calculated solid-phase loading expected with an influent cVOC concentration of 5 μg/L. Except where noted, Freundlich K and 1/n values and equilibrium concentrations are from Speth and Miltner (1990).

Compound	MCL (μg/L)	MDL (ng/L)	LCMRL (ng/L)	KF (μg/g)(L/μg)1/n	1/n (−)	Min Ce (μg/L)	q5 - solid phase loading (μg/g)
Vinyl Chloride#* (VC)	2	3.2	11	0.73	0.340		1.26
Dichloromethane (DCM)	NR	11	116	6.25	0.801	18.1	22.7
1,1,2,2-Tetrachloroethane (1,1,2,2 TeCA)*	NR	7.1	30	69.7	0.370		126
1,1-Dichloroethane (1,1 DCA)	NR	16	23	65	0.706	13.4	201
1,2-Dichloroethane (1,2 DCA)	5	7.9	20	129	0.533	42.8	304
1,2-Dichloropropane (1,2 DCP)	NR	16	25	313	0.597	4.1	818
Carbon Tetrachloride (CT)	5	21	21	387	0.594	9.1	1007
1,3-Butadiene#*	NR	16	43	771	0.520		1780
1,1,1,2-Tetrachloroethane (1,1,1,2 TeCA)	NR	16	16	1070	0.604	1.2	2829
1,2,3-Trichloropropane (1,2,3 TCP)	NR	20	37	1080	0.613	3.4	2897
Benzene	5	23	24	1260	0.533	3.2	2948
Trichloroethylene (TCE)	5	10	17	2000	0.482	7.7	4358
Tetrachloroethylene (PCE)	NR	30	30	4050	0.516	3.6	9292

NR- not regulated

^#Analyzed using SPME

^*Freundlich K and 1/n calculated from AdDesignS

Sample Analysis

Concentrations of VOCs were analyzed via EPA method 524.3. However, the method was modified to use heated headspace analysis instead of purge and trap for cVOCs that are liquid at room temperature. For the gaseous VOCs (VC and 1,3-butadiene), the method used solid-phase micro-extraction (SPME) analysis. In addition, the method used an MS detector in the single ion mode (SIM) instead of full scan mode to obtain lower method detection limits (MDL)¹. The lowest concentration minimum reporting levels (LCMRLs)² were determined in DI water. The LCMRLs were determined between 10 – 50 ng/L, with the exception of dichloromethane (Table 2).

Samples were preserved by pH adjustment to 2 with either maleic acid (OH) or hydrochloric acid (CO I and II). Prior to analysis, a 20-mL amber headspace vial was filled with 15-mL sample. A gram of NaCl was added and the vial was heated to 60 °C for 15 minutes in an autosampler agitator³ connected to a GC/MS⁴. For cVOCs that are liquids at room temperature, a 2500-μL gas sample was injected into the GC/MS equipped with a split/splitless injection port operated in the splitless mode. For chemicals that are gases at room temperature, a SPME fiber connected to the autosampler was exposed to the headspace for 5 minutes, and the fiber was desorbed for 5 minutes in the injection port of the GC at 250 °C.

TOC samples were collected in 40-mL vials and analyzed according to EPA method 415.3 with a high temperature combustion catalytic TOC analyzer⁵.

Carbon Preparation

The GAC was a fresh bituminous-based coal typical for VOC adsorption⁶. A 50-lb sample was reduced to a manageable volume via the method of coning and quartering (USEPA, 1996). The entire sub-volume was carefully crushed with a mortar and pestle and separated with U.S. Standard sieves on a sieve shaker. The fraction passing #100 sieve but retained on #200 sieve (log mean particle diameter = 0.11 mm) was collected for the experiment. The collected fraction was repeatedly washed and decanted to remove excess fines, then dried overnight in an oven at 105 °C and stored in a desiccator.

Experimental Design

RSSCTs were designed according to the proportional diffusivity (PD) approach because the PD RSSCT has been shown to better predict breakthrough when DOM is present (Crittenden et al., 1991, Corwin and Summers, 2010). The GAC was packed in 1/4 in. Teflon® columns, providing a column diameter to particle size ratio greater than 34, large enough to avoid wall effects (USEPA 1996). The design of the full scale column was based on a 15 min EBCT, resulting in a scaled RSSCT EBCT of 1.76 min. Scaling down to the RSSCT, the RSSCT flow rate was 1.4 mL/min (OH) or 2.0 mL/min (CO), set by the minimum allowable Reynolds number (Re_min) for DCM because DCM has the highest diffusivity in water leading to the lowest Schmidt number (Sc) (Re_min=500/Sc) (Crittenden et al., 1987). The small RSSCT flowrate differences for OH and CO experiments resulted from small differences in the internal diameter of the nominal 1/4 in. Teflon® columns.

cVOC laden water was pumped from the container (PFA bag or glass carboy with volatile trap) to the GAC column with either a ceramic/fluorocarbon-PVDF piston pump⁷ (OH) or a PTFE diaphragm pump⁸ (CO) mated to a variable speed (1-100 rpm) console drive⁹. Tubing was 1/8 in. – 1/4 in. Teflon® or stainless steel with stainless steel connectors¹⁰. A glass wool pre-filter was placed before the adsorbent column and was changed out weekly to avoid biological growth. A pulse dampener¹¹ with only stainless steel and Teflon wetted parts was inserted between the pump and the GAC columns. Two equal columns in series were used to create a 7.5 minute and a 15 minute GAC EBCT (full-scale equivalent). Before the first column, in between the two columns, and after the second column were stainless steel three-way valves¹⁰ for sampling. Sampling was performed by attaching a 50 mL glass syringe¹² to a Luer-Lock connecter on the valve and allowing it to fill headspace free. Once full, the contents of the sample syringe were ejected into 40-mL sample vials containing the preservative and sealed with Teflon faced septa, stored at 4°C, and analyzed within 14 days. The treatment goal for all experiments was 0.5 μg/L.

Results and Discussion

RSSCT results.

The breakthrough results in the OH GW are shown in Figures 1a and b at EBCTs of 7.5 and 15 minutes, respectively. The results are reported as normalized effluent concentration (the ratio of the effluent concentration, C, to the influent concentration, C₀) as a function of throughput in bed volumes, which is volume of water treated normalized by GAC bed volume. The average influent concentration and associated coefficient of variation are reported in Table 3 for the RSSCTs, from the shortest column run time in bed volumes to 10% breakthrough (DCM - 1.25x10³) to the longest (PCE - 330 ×10³) (except for 1,2 DCA-CO II GW). The benzene breakthrough data are not shown because benzene concentrations decreased rapidly in the influent. The likely cause for benzene degradation was biological activity (Kim, 2003) and although trials with sodium azide as an inhibitor appeared promising (data not shown), resources did not allow for benzene RSSCTs to be continued. Additionally, the 1,1,2,2 TeCA data are not shown because the experimental RSSCT influent was contaminated with TCE, preventing further analysis of the data.

Figure 1.

Breakthrough curves for 11 cVOCs in OH GW (1.0 mg/L TOC) a) 7.5 min EBCT and b) 15 min EBCT. Expanded view of 9 cVOCs in OH GW; c) 7.5 min EBCT and d) 15 min EBCT. Dashed black line represents a typical TOC breakthrough curve for the OH GW. Light dashed horizontal line represents a C/C₀ value of 1.0. VC breakthrough was not assessed at 15 min EBCT.

Table 3 –

Summary of breakthrough results at 7.5 min EBCT – Capacity for cVOCs measured in thousands of bed volumes treated to reach 10%, 50%, and 75% breakthrough, the throughput of bed volumes to peak concentration (BVP), the normalized peak effluent concentrations, and the capacity differences between 7.5 and 15 min EBCTs are shown.

Compound	Water	Average C0 (μg/L)	Coeff. of Variation (%)	Throughput (103 Bed volumes) to 10%, 50%, and 75% Breakthrough			Throughput (103 Bed volumes) to peak conc.	Peak C/C0	Capacity difference 7.5 min vs 15 min EBCT* (%)
				BV10	BV50	BV75	BVP		BV10	BV50	BV75
DCM	OH	4.1	5	1.25	1.55	1.70	3.05	1.36	−3	5	6
VC	OH	7.7	2	2.61	3.40	3.90	5.86	1.10	na	na	na
1,1 DCA	OH	4	14	14.8	20.0	22.7	37.3	1.25	−5	3	13
1,2 DCA	OH	5.6	12	18.5	25.4	26.5	38.8	1.52	−17	8	11
1,2 DCA	CO I	4.6	34	20.6	23.0	23.2	44.8	1.45	−3	−4	−5
1,2 DCA	CO II	3.9	9	14.5	16.2	16.7	39.5	1.27	8	14	15
1,2 DCP	OH	4.7	13	55.0	66.8	69.1	132	1.59	−9	4	6
1,3 Butadiene	OH	4.1	51	59.0	62.0	63.5	74.1	1.83	−6	−7	−7
1,1,1,2 TeCA	OH	5.5	26	78.0	96.8	105	168	1.95	4	19	18
CT	OH	4	29	83.0	105	119	163	1.80	0	21	31
1,2,3 TCP	OH	7.3	21	107	124	137	156	1.80	5	12	21
TCE	OH	4.2	33	178	198	204	291	1.42	19	21	22
PCE	OH	2.7	49	330	362	388	459	2.32	6	14	19
TOC	OH	1.0 mg/L		-	25.0	-	-	-	na	na	na

^*capacity difference = [(throughput_7.5min – throughput_15min)/ throughput_15min] x 100%;

To better observe the breakthrough behavior of less strongly adsorbing compounds, TCE and PCE results were removed from the plot and the breakthrough data plotted to only 150,000 bed volumes in Figures 1c and d, for 7.5 and 15 minute EBCTs, respectively. A typical TOC breakthrough for the OH GW is also plotted as a heavy dashed line in Figures 1c and d. For reference, 100,000 bed volumes represents about 1.5 and 3.0 years of full scale operation at EBCTs of 7.5 and 15 minutes, respectively. For the most part, breakthrough of cVOCs from the column occurred in accordance with their Freundlich values, which have units of (μg/g)(L/μg)^1/n. DCM and CT were exceptions. DCM (K_F = 6.25) reached breakthrough before VC (K_F = 0.73), although with such low K_F values for both chemicals, they both reached breakthrough in just a few thousand bed volumes. Based on K_F, CT (K_F = 387) should have reached breakthrough closer to 1,2 DCP (K_F = 313); however, breakthrough occurred later than 1,1,1,2 TeCA (K_F = 1070). Finally, it’s worthwhile to observe that substantial TOC breakthrough occurs prior to most of the cVOCs.

The cVOC RSSCT breakthrough results at an EBCT of 7.5 minutes, and with 1,2 DCA in three waters, are summarized in Table 3, including the throughput in bed volumes treated to 10%, 50%, and 75% breakthrough (BV10, BV50, and BV75, respectively). The throughput to a given breakthrough at an EBCT of 7.5 minutes was found to be related to the equilibrium solid-phase loading, q₅, calculated using the Freundlich isotherm values from Table 2 and a C₀ value of 5 μg/L, the target influent VOC concentration. These equilibrium isotherms were run as single solutes in distilled water with a bituminous based GAC (Speth and Miltner, 1990). The minimum equilibrium concentrations from the isotherms conducted by Speth and Miltner (Figure 2) range between 1.2 and 42.8 ug/L, which is a similar magnitude to the concentrations in this study such that extrapolation effects should be minimal. Previous work by Jarvie et al. (2005) also used single solutes to develop empirical equations predicting adsorbent capacity reduction. Jarvie’s work considered characteristics of the water, the pre-loading and competitive contributions of DOM and the target organic. Other efforts have explored competitive effects to adsorption using single solutes and ideal adsorbed solution theory (Crittenden et al., 1985). The results presented here are a simplified effort to provide linear relationships between the BV10 and BV50 values in OH GW and q₅ values and are shown in Figure 2. The relationship found is encouraging because it implies there may be a method to estimate the throughput to a set point of breakthrough from single solute isotherms using the same GAC, which may be extended to additional waters if the effects of DOM can be quantified. For example, Kempisty and Summers (2016) developed relationships for throughput to 10% breakthrough for 1,2 DCA in different waters and with different TOC concentrations and characteristics. Using a regression model focusing on the influent 1,2 DCA concentration, ultraviolet absorbance at 254 nm (UV₂₅₄) as proxy for TOC concentration, and the fluorescence index as a measure of microbial and terrestrial characteristics of the DOM, the relationship predicts 20,300 bed volumes for the OH GW while the observed value was 18,500 bed volumes – less than a 10% difference. While the relationship between bed volumes treated and q₅ is encouraging, a t-test to compare slopes of regression lines resulted in a p-value (0.39) greater than alpha (0.05) indicating the slopes of the BV10 and BV50 relationships are not statistically different.

Figure 2.

Relationship between calculated solid-phase loading, q₅ (and influent concentration 5 μg/L), and throughput to 10 or 50% breakthrough at 7.5 min EBCT for 11 cVOCs (VC, DCM, 1,1 DCA, 1,2 DCA, 1,2 DCP, 1,3 Butadiene, 1,1,1,2 TeCA, CT, 1,2,3 TCP, TCE, PCE) in OH GW.

Displacement desorption.

One indicator of competition between cVOCs and other adsorbates (e.g., DOM) is the observation of effluent concentrations from a GAC adsorber exceeding influent concentrations. This phenomenon, called displacement, can occur when competing adsorbates (either DOM or another target organic) are introduced into an adsorber. A weaker adsorbing compound will travel through an adsorber faster than a more strongly adsorbing compound. When the stronger compound reaches further into the GAC bed, it will compete with the (previously adsorbed) weaker compound for an adsorption site and may displace the weaker compound. This leads to an effluent containing both the mass of the contaminant in the influent and the mass of the displaced contaminant and results in a normalized effluent concentration (C/C₀) greater than 1.0. This phenomenon is also called the chromatographic effect and is shown in Figure 1 where the normalized effluent concentration eventually exceeds a value of 1.0 for all VOC runs at both EBCTs.

The normalized peak effluent concentration and corresponding throughput are also included in Table 3 for 7.5 minute EBCT. The displacement desorption resulted in peak C/C₀ values ranging from 1.10 for VC to 2.32 for PCE. This has been observed in other work but not at these concentrations or with this variety of compounds (Summers et al. 1989). Since each column was operated with a single VOC, the competitive organic matter is attributed to the DOM in the background matrix of the OH GW. Understanding the effects of displacement desorption is important because influent concentrations may be below a treatment objective while effluent concentrations may exceed that threshold.

An example of large displacement desorption can be seen in Figure 3 for 1,2,3 TCP. The 7.5 minute EBCT column resulted in a large peak C/C₀ and an effluent concentration of 13 μg/L, nearly twice the influent concentration because of displacement desorption. The 15 minute EBCT column was discontinued shortly after 100% breakthrough and therefore the peak displacement concentration was not captured. While the DOM breakthrough, measured by TOC, reached 75% after approximately 65,000 bed volumes, a small well-adsorbed fraction of the DOM was likely still being removed, even after 150,000+ bed volumes as evidenced by the TOC C/C₀. The well adsorbed DOM fraction competes with the sorbed VOC and causes displacement. Even if this well adsorbed DOM fraction was only 5% of the influent TOC, the concentration would be 50 μg/L, which is 10 times as high as that of the VOCs. Sontheimer et al. (1988) discuss the distribution of adsorbable DOM fractions and show about 10% of a typical source water DOM has higher Freundlich K values than TCE or PCE, the highest adsorbing VOCs evaluated in this study. Generally, the peak effluent concentration increases with the adsorbability of the VOC, because the greater equilibrium solid-phase loading results in more VOC for displacement.

Figure 3.

Breakthrough curves of 1,2,3 TCP (C₀ = 7.3 μg/L) and TOC (C₀ ~ 1.0 mg/L) at two EBCTS, 7.5 and 15 min, from OH GW.

GAC fouling.

The impact of EBCT on breakthrough is illustrated in Figure 3 for 1,2,3 TCP. While the beginning of the 1,2,3 TCP breakthrough for the 7.5 minute and 15 minute EBCTs were similar, after 100,000 bed volumes the 15 minute EBCT resulted in earlier breakthrough compared to that of the 7.5 minute EBCT. The capacity reduction at longer EBCT may be a result of GAC fouling which occurs because of more strongly adsorbing and non-displaceable DOM reaching deeper in the adsorber bed and outcompeting the target organic for adsorption sites and/or covering the pores in the GAC particles over time and preventing further target organic adsorption. The impacts of EBCT on GAC performance are illustrated for the other cVOCs in a comparison of the capacity difference in throughput at similar breakthrough points (BV10, BV50, and BV75) (Table 3), where negative values indicate greater capacity at 15 minute EBCT and positive values indicate greater capacity at 7.5 minute EBCT (or lesser capacity at 15 minute EBCT). The average capacity difference (excluding VC because 15 minute EBCT was not assessed) measured in number of bed volumes to 10%, 50% or 75% breakthrough was −0.1%, 9% and 13%, respectively, indicating no loss of capacity at 10% breakthrough and a modest loss of capacity with the 15 minute EBCT at 50% and 75% breakthrough because of fouling at longer EBCTs.

Depending upon the nature of the organic matter in the background matrix of water being treated, increasing EBCT may result in more or less adsorption of target organics because of co-loading or preloading conditions and DOM fouling (Kempisty and Summers, 2016, Knappe et al., 1997). The effect of EBCT on both the extent of column exhaustion (BV10 vs. BV50 vs. BV75) and the affinity of the VOC to the GAC (throughput measured in bed volumes) is shown in Figure 4. In the figure, moving from negative to positive percent difference illustrates when the 15 minute EBCT was less effective because of fouling. The dashed line represents the average (n = 20) bed volumes to 50% TOC breakthrough in the OH GW; 23,000 bed volumes. At throughput values similar to those of TOC breakthrough, some of the BV10 values were negative and more favorable in the 15 minute EBCT, but their corresponding BV50 and BV75 values were positive and more favorable in the 7.5 minute EBCT, indicating some fouling as the column exhaustion percent increases.

Figure 4.

Effect of EBCT on throughput to 10%, 50% and 75% breakthrough; dashed line reflects average BV50 TOC throughput (23×10³ bed volumes) for n = 20 experiments conducted in OH GW. ** percent difference in throughput = [(throughput_7.5min – throughput_15min)/throughput_15min] x 100%;

As column capacity (and operating time) increases for the more adsorbable cVOCs, the DOM in the background matrix, having a faster mass transfer zone than the target cVOC, travels ahead of the cVOC and reduces the GAC capacity in columns with greater EBCTs. As shown in Figure 4, adsorbers operating for longer periods of time (both greater throughput and greater extent of column exhaustion) have greater cVOC capacity at EBCTs of 7.5 minutes versus 15 minutes as indicated by the general shift of data to the right as the throughput increases. In 16 of 17 cases, 15 minute EBCT columns operating longer than 66,000 bed volumes had reduced capacity and increased DOM fouling compared to their 7.5 minute counterparts (in the one outlier case, CT capacity difference for BV10 was 0). Previous work supports this observation. Summers et al. (1989) suggested greater DOM-adsorbent contact time was responsible for greater fouling in the adsorbers, while Hand et al. (1989) suggested reducing the ratio of adsorbate mass transfer zone to column length would cause lower relative bed volumes fed (or CUR) for longer columns at high percent breakthroughs. Shih et al. (2003) observed little effect of EBCT on methyl tert-butyl ether breakthrough in RSSCT experiments involving groundwaters (TOCs: 1.0 and 0.2mg/L) but decreased adsorption capacity when EBCTs were increased from 10 to 20 minutes with a surface water (TOC: 3.2mg/L). The groundwaters in the current effort generally resulted in decreased adsorption capacity at the higher EBCT as the TOC increased from 0.3 mg/L (CO I) to 1.5 mg/L (CO II).

Effects of varying background matrix.

Using 1,2 DCA as a model adsorbate, RSSCTs were conducted in three different water sources: OH GW, CO I GW, and CO II GW. The three water sources were chosen for their varying DOM content, low CO I GW (0.3 mg/L TOC), mid-range baseline OH GW (1.0 mg/L TOC), and high CO II GW (1.5 mg/L TOC). The 1,2 DCA breakthrough results for these three waters are shown in Figure 5a and b for both the 7.5 minute and 15 minute EBCTs, respectively. At both EBCTs, 1,2 DCA breakthrough occurs about 30% earlier in the CO II GW, the water with the highest influent TOC. The reduction in target compound performance at higher influent TOC concentrations has been reported and modeled for 1,2 DCA (Kempisty and Summers, 2016) and other target compounds (Corwin and Summers, 2012; Summers et al., 2013, Kennedy et al., 2015). In contrast, comparing the CO I and OH GWs did not result in significant differences. The 15 minute EBCT breakthrough showed little difference between the two waters. The 7.5 minute EBCT resulted in some difference near the middle of the 1,2 DCA breakthrough although it is not enough overall to demonstrate adsorption differences due to the varying DOM of the two waters, even though it was expected that the low TOC GW (CO I) would yield the best performance for 1,2 DCA. Although the throughput differences between the two waters were minimal, a model developed using the CO I and CO II GWs for 1,2 DCA (Kempisty and Summers, 2016) resulted in less than 10% difference between actual and predicted bed volumes for the OH GW, as described earlier in the manuscript. Therefore, it is likely that factors other than the TOC concentration may be responsible for these results. High molecular weight DOM (steric hinderance or external film mass transfer effects), the humic or fulvic content, and the age of the DOM can all affect carbon-adsorbate-DOM interactions. Additionally, DOM is heterogenous in nature and can vary temporally causing effects not predictable by TOC concentration alone. Although the results are for only one VOC and three waters, it demonstrates the significant role DOM concentration and character can contribute to the carbon efficiency in removing target organic compounds.

Figure 5.

Effect of background matrix on 1,2 DCA adsorption at EBCTs of a) 7.5 min and b) 15 min.

Carbon use rates (CUR).

CUR is a measure of the mass of carbon required to treat a volume of water to a desired concentration and was calculated using Equation 1 for the VOC breakthroughs at both EBCTs. The results are provided in Table 4. The GAC bed density was 0.48 g/cm³ and the treatment objective was 0.5 μg/L.

Table 4.

Carbon Use Rates for cVOCs at 2 EBCTs. All results are for OH GW with the exception of 1,2 DCA which are also shown for CO I and CO II waters. Treatment objective was 0.5 μg/L.

Compound	Carbon Use Rate (lb GAC/1000 gal treated)
	7.5 min EBCT	15 min EBCT
DCM	3.1	3.0
VC	1.6	na
1,1 DCA	0.26	0.23
1,2 DCA	0.22	0.18
1,2 DCA (CO I)	0.19	0.19
1,2 DCA (CO II)	0.27	0.30
1,2 DCP	0.073	0.067
1,3 Butadiene	0.068	0.064
1,1,1,2 TeCA	0.051	0.054
CT	0.048	0.048
1,2,3 TCP	0.037	0.039
TCE	0.023	0.027
PCE	0.012	0.013

$$CUR=\frac{GACbeddensity}{\#bedvolumestreatedtotreatmentobjective}$$ (1)

From an operations perspective, lower CURs are desirable because they lead to less carbon required to remove a given chemical. In this research, a CUR of 0.2 pounds GAC per 1000 gallons water treated (~24 mg GAC per liter water treated) was considered an indicator of GAC treatment feasibility. The selected 0.2 CUR is not an absolute feasibility value but rather a subjective balance between effectively removing cVOCs from water without excessively changing carbon and resulted from discussions among water treatment professionals. Values significantly above this threshold call into question the viability of GAC as an effective treatment technology because of short GAC replacement intervals. Based on this threshold, and for the specific waters, GAC, and low concentration treatment objective used in this research, GAC is viable for the treatment of 7 of the VOCs investigated in this effort. GAC is not cost effective for the removal of VC or DCM because of their fast breakthrough and excessively high CUR (>1.6 lbs/1000gal). Two VOCs, 1,1 DCA and 1,2 DCA, have CURs near the approximate threshold value for viability and may be feasible under specific water quality conditions although further testing would be recommended. However, the research conducted herein only evaluated single solutes as opposed to mixtures of contaminants. Such mixtures of cVOCs, depending upon their adsorbability, could negatively affect the CURs.

The CURs at an EBCT of 7.5 min for the OH GW are plotted as a function of equilibrium solid-phase loading (q₅) and a power function relationship fit to the RSSCT data with 0.92 R² (n=9) (Figure 6). CURs obtained using the RSSCT require caution be used because RSSCTs tend to over-predict full-scale breakthrough times and under predict CURs. Kennedy et al. (2015) developed a scaling relationship between RSSCTs and pilot-scale adsorbers for throughput to 10% breakthrough for more than 30 compounds including some VOCs:

$$\text{ln}B{V}_{10\%}^{pilot}=0.57+0.855\text{ln}B{V}_{10\%}^{RSSCT}$$ (2)

Figure 6.

Relationship between CUR for OH GW RSSCT and for predicted pilot-scale at 7.5 min EBCT with calculated solid-phase loading, q₅ (and influent concentration 5 μg/L). Solid regression line represents RSSCT data; dotted regression is predicted pilot-scale data; dashed line is CUR of 0.2 lbs/1000 gal.

The calculated CUR values representative of the pilot-scale, which is similar to full-scale, are also plotted in Figure 6 for the same VOCs as the RSSCT data. A similar power function fit the relationship between q₅ and the pilot-scale estimated CURs. The CURs estimated for the pilot-scale adsorbers were about 3 times as high on average than that calculated with the RSSCTs (Kennedy et al. 2015). Using this scaling relationship, the pilot-scale CURs for 1,1 DCA and 1,2 DCA were above 0.5 lbs/1000 gal, while the RSSCT CURs were at the approximate threshold value for viability. Therefore, viability of GAC as a treatment technology for 1,1 DCA and 1,2 DCA to a treatment level of 0.5 μg/L is unlikely. Since 1,1 DCA and 1,2 DCA are so close to the feasibility limit, it may be possible to use these cVOCs as a reference for viability of other cVOCs. The scaling exercise demonstrates that to obtain better CUR estimates for full-scale adsorbers, pilot-scale adsorbers may best be utilized to determine appropriate correction factors applicable to the particular water of interest, in addition to RSSCT evaluations.

Conclusions

Twenty-eight RSSCTs were operated at two EBCTs of 7.5 minutes and 15 minutes to assess the GAC performance for 13 VOCs at low μg/L feed concentrations with a target effluent concentration of 0.5 μg/L in groundwater. One compound, 1,2 DCA, was also evaluated in two other groundwaters. VC and DCM reached breakthrough very quickly, indicating that GAC is not a viable technology. For the remaining VOCs, RSSCTs resulted in a wide range of breakthrough times from 14.8 thousand bed volumes for 1,1 DCA to 330 thousand bed volumes for PCE. The throughput to 10% and 50% breakthrough in the OH GW was linearly correlated to capacities calculated with single-solute equilibrium isotherm parameters (R² = 0.97, 0.96, respectively). All VOC compounds successfully studied in RSSCTs, including VC and DCM, resulted in displacement desorption with effluent concentrations exceeding the influent concentrations. This was attributed to the strong adsorbing DOM fraction in the groundwater, as all runs were conducted with single VOCs spiked into the groundwater. Evidence of a background DOM fouling effect was found, leading to modest decreases in throughput, 9 to 13% on average to 50% and 75% breakthrough when the EBCT was increased from 7.5 min to 15 min. The difference in breakthrough between the two EBCTs increased with the more strongly adsorbing compounds. CUR, when scaled to simulate the full-scale adsorber, indicated that GAC would be a viable technology for seven of the VOCs successfully evaluated when using a CUR threshold less than 0.2 lbs/1000 gal. Since 1,1 DCA and 1,2 DCA are so close to the feasibility limit, it may be possible to use these cVOCs as a reference for viability of other cVOCs. The impact of the background DOM and GAC type are not well understood and until additional research elucidates this behavior, additional evaluation in the particular water matrix with the GAC of interest is recommended for compounds with CURs near 0.2 lbs/1000 gals.