This article describes a cost-effective and sensitive approach for quantifying waterborne arsenic based on gas-phase chemiluminescence. The approach centers on the use of photometric instruments-one configured for laboratory use and one field-deployable version-that can quantify total arsenic as well as individually measure As(III) and As(V). The regulatory limit for arsenic in drinking water is 10 μg/L. The limits of detection of the gas-phase chemiluminescence instruments are well below 1 μg As/L and the linear range extends to >100 μg As/L. Total arsenic analysis using this approach requires 3 min.

	
	Arsenic in drinking water is a serious issue in many parts of the world (1). Arsenic is a Class A human carcinogen, and it can lead to skin, bladder, lung, and prostate cancer (2). It has also been linked with respiratory, reproductive, developmental, immunological, and neurological defects (2).
	
	The maximum permissible level of arsenic in drinking water, according to the World Health Organization as well as the United States Environmental Protection Agency (US EPA), is 10 μg/L. However, such a standard is presently too stringent to be met in many places in the world; in India, Bangladesh, Taiwan, China, and Vietnam, the current limit is 50 μg/L (3). Arsenic is the 20th most abundant crustal element; the presence of arsenic in groundwater is common. The Trace Elements National Synthesis Project of the United States Geological Survey has created a map of groundwater arsenic distribution in the United States (4); high (50 μg/L) levels of arsenic in groundwater exist in many locations. A recent publication has also reported increased occurrence of elevated arsenic levels in groundwater after local hydraulic fracturing activities in the Barnett Shale area of Texas (5).

	Present US EPA-approved detection methods are all based on atomic spectrometry, and “approved” methods must be used for regulatory reporting. For non-regulatory testing, any method may be used, but it is beneficial if the method or instrument performance has been verified by the US EPA Environmental Technology Verification program. One stripping voltammetric analyzer has undergone such verification, but electrochemical analyzers often suffer from electrode fouling, making standard addition a necessity; even traces of copper present can be a problem (6). Other reports have indicated the high degree of expertise needed to obtain reliable results with electrochemical analyzers (7). On the other hand, atomic spectrometry–based instruments are bulky, expensive and require large amounts of pure gas in addition to expensive consumables. Moreover, such instruments cannot be used in the field.

	Field analysis is extremely important considering the temporal changes in arsenic concentrations at various sampling sites and the uncertainty that always surrounds the integrity of a sample preserved in the field. Currently commercially available field assays are based on the Gutzeit method (8), which involves generation of arsine (AsH

), filtering through lead acetate paper to trap any hydrogen sulfide formed, and capturing the arsine on a mercuric bromide-soaked paper to induce a yellow coloration. Arsine that is produced can escape from the device, thus posing a health hazard (9). The less expensive instruments use comparison with a color chart for stepwise quantification; color degradation in sunlight and operator judgment can be factors (10). Undesirable aspects of methods involving those instruments include the use of large sample volumes, a corresponding amount of acid, and lead and mercury compounds that generate toxic waste.

	A field-usable instrument must be robust and easy to operate. Other important factors are reproducibility, capital and consumable cost, the ability to differentiate inorganic As(III) and As(V) (which differ considerably in their ease of removal and acute toxicity), and environmental friendliness. In the past, work in our laboratory demonstrated the principles of a gas-phase chemiluminescence-based arsenic analyzer (11). This article describes the use of prototype laboratory-based and field-deployable gas-phase chemiluminescence analyzers developed in our laboratory to perform simple, fast, environment-friendly, and fully automated analysis of waterborne arsenic. The instruments measure total arsenic or As(III) (with As(V) by difference), or As(III) and As(V) sequentially. Other configurations currently in development in our laboratory include in-line continuous process instruments and remotely deployed autonomous instruments. 
	
	


	
	Standard 1000-mg/L stock solutions of As(III) and As(V) were prepared using arsenic trioxide and sodium arsenate heptahydrate, respectively. Working solutions were prepared by serial dilution with deionized water (18 MΩ cm

) generally immediately before use. For hydride generation, 4.8% (w/v) sodium borohydride (NaBH

) (98%) was prepared in 0.5 M sodium hydroxide and 1 mM disodium ethylenediamine tetraacetate (Na

EDTA). For sequential analysis of As(III) and As(V), citrate buffer (pH 4.5) was used. Citrate buffer (pH 4.5) was prepared by adding sodium hydroxide pellets to 1 M citric acid and then making fine adjustments with 2 M sodium hydroxide. In this approach, 1 mL of citrate buffer is added to the sample followed by 0.5 mL of the sodium borohydride reagent. For the subsequent As(V) measurement, another 0.5 mL aliquot of sodium borohydride is added, followed by 1 mL of 6.0 M sulfuric acid (H


	
	A basic schematic of the instrument is shown in Figure 1. Figure 2 shows an inside view of the laboratory instrument. The laboratory and field-deployable instruments both are fully automated. All liquid handling is conducted by a syringe pump (designated as “SP” in Figure 1) connected to a multiport distribution valve (DV). The laboratory setup has an eight-port distribution valve, and the portable instrument has a six-port distribution valve. The distribution valve ports are respectively connected to a sample container, a waste bottle, a reaction chamber (RC), and reservoirs respectively containing solutions of sodium borohydride, sulfuric acid, citrate buffer, and water. One port of the distribution valve in the laboratory instrument is left unused and vented to the atmosphere. With the field instrument, one of the ports of the six-port distribution valve is connected to the common port of a three-way solenoid valve to effectively create a seven-position valve. The normally open and normally closed ends of the solenoid valve are connected to reservoirs containing solutions of sulfuric acid and citrate buffer, respectively. The temporal operational sequence, aspiration-dispense volumes, and aspiration-dispense velocities are programmable. The reaction chamber (RC) is a 20-mL cylindrical chamber made of poly(methyl methacrylate) (PMMA), the bottom of which is connected to a tube that drains to waste via a normally closed solenoid valve (SV1). Three fluid lines enter through the top of the reaction chamber. One of the tubes reaches the bottom of the reaction chamber and carries all the liquid delivered by the syringe pump. A second line reaches to a point inside the reaction chamber that is just above the maximum fill volume of the reaction chamber in operation such that the headspace can be swept. It forms an air delivery line that comes from a three-way solenoid valve (SV2), and passes through a flow restriction tube (R2) (PTFE, 0.30 mm i.d., 68 cm long). The third fluid line entering the reaction chamber is a gas exit line and terminates at the top. It goes to a third solenoid valve (SV3), identical to SV2. The flow restrictors, R1 and R2, join at a tee (T), which splits the output of an air pump. The other side of the tee goes through restrictor tube R1 (PTFE, 0.30-mm i.d., 34 cm long) and through an ozone generator. The ozone generator consists of two concentric glass tubes placed inside a polypropylene tee. The inner glass tube (0.7 mm i.d., 2 mm o.d., 16 cm long) is sealed at both ends and contains a nichrome wire (0.5 mm thick, 16.5 cm long) that serves as the high-voltage anode. The outer surface of the other glass tube (4 mm i.d., 6.2 mm o.d., 10 cm long) is wrapped with aluminum foil, and it serves as a cathode. The optimized flow rates were 60 mL/min through ozone generator and 30 mL/min for the headspace flush flow. The output of the ozone generator and the output line of the reaction chamber from the third solenoid valve (SV3) terminate in a polypropylene tee on top of the chemiluminescence chamber (CC). The chemiluminescence chamber is fabricated from the bottom ~5-cm portion of a 12-mm i.d. glass test tube. It is silvered on the exterior with a commercial silvering solution and then painted black repeatedly with black epoxy-based paint.

	
	The bottom of the test tube is closed by covering it with a 0.15-mm-thick, 25-mm-square microscope cover glass. A hole (3 mm) is drilled into the top of the dome. The tee assembly is then cemented using epoxy adhesive. The tee assembly has a glass tube (1.7 mm i.d., 3.0 mm o.d., 4.6 cm long) attached to it at the bottom, and it carries both ozone and arsine to the bottom of the chemiluminescence chamber so that they have sufficient time to react and produce chemiluminescence. The length of the PTFE tube carrying arsine inside the chemiluminescence chamber can be changed by either pulling it out or pushing it inside the glass tube. Another hole (1 mm) is drilled near the top to serve as the gas exit and it leads to a vent cartridge containing granular activated charcoal to catalytically destroy any excess ozone in the exit stream. The volume of the chemiluminescence cell is ~5.5 cm

. All the conduits leaving and entering the chemiluminescence chamber were opaque black PTFE tubes (1.1-mm i.d.). The chemiluminescence chamber is mounted on top of a miniature photomultiplier tube (PMT). The PMT along with the chemiluminescence chamber is placed in a plastic opaque black box so that there is no interference from room light. The output of the PMT was further amplified using a secondary two-stage operational amplifier. The overall gain is 1000 × with a time constant of 1 s.

	For the laboratory instrument, the data were acquired on a 12-bit USB data acquisition card using LabVIEW (National Instruments)-based software developed in-house. The LabVIEW program also triggers the operational sequence of the syringe pump and the opening of the of the solenoid valves through the digital outputs of the same data acquisition card. The final results are displayed on a laptop computer. The operational sequence of the portable instrument was controlled by a microcontroller and the results were displayed on a color multitouch liquid crystal display panel.


	
	The typical operational sequence for total arsenic estimation for both instruments is similar, as follows:
	1. Aspirate 2.5 mL of sample and dispense it to the reaction chamber.
	2. Aspirate 0.5 mL of sulfuric acid and dispense it to the reaction chamber.
	3. Aspirate 1 mL of water and dispense it to the waste reservoir. Repeat this step four times to rinse and clean the syringe.
	4 Aspirate 0.5 mL of sodium borohydride solution and dispense all of it to the reaction chamber.
	5. Wait for 30 s to allow the arsine gas to accumulate at the headspace of the reaction chamber.
	6. Open the solenoid valve SV3.
	7. After 2 s, open the solenoid valve SV2 and record the signal height for 60 s. This step allows the air to purge all the arsine gas out of the reaction chamber.
	8. Close solenoid valve SV3 and open solenoid valve SV1 to allow the contents of the reaction chamber to drain off.
	9. Close solenoid valves SV1 and SV2.
	10. Aspirate 2.5 mL of water and dispense it to the reaction chamber. Repeat this step one more time so that the reaction chamber is flushed with 5 mL of water.
	11. Open solenoid valves SV1 and SV2 so that the contents are drained out.
	12 Close solenoid valves SV1 and SV2.
	
	The software converts the recorded signal height into the amount of the total arsenic using a calibration curve and displays it on-screen. The complete cycle takes ~3 min. For the laboratory setup, the operational protocol for the sequential analysis of As(III) and As(V) is as follows:
	
	1. Aspirate 2.5 mL of sample and dispense it to the reaction chamber.
	2. Aspirate 1 mL of citrate buffer and dispense it to the reaction chamber.
	3. Aspirate 1 mL of water and dispense it to the waste reservoir. Repeat this step four times to rinse and clean the syringe.
	4. Aspirate 0.5 mL of sodium borohydride solution and dispense all to the reaction chamber.
	5. Wait for 30 s to allow the arsine gas to accumulate at the headspace of the reaction chamber.
	6. Open the solenoid valve SV3.
	7. After 2 s, open the solenoid valve SV2 and record the signal height for 60 s. This measurement gives the signal height for As(III) species.
	8. Aspirate 1 mL of water and dispense to the waste. Repeat this step four times to rinse the syringe.
	9. Aspirate 1 mL of sulfuric acid and dispense to the reaction chamber.
	10. Aspirate 1 mL of water and dispense it to the waste. Repeat this step four times.
	11. Aspirate 0.5 mL of sodium borohydride solution and dispense all to the reaction chamber.
	12. Wait for 30 s to allow the arsine gas to accumulate.
	13. Open solenoid valve SV3.
	14. After 2 s open solenoid valve SV2 and record the signal height for 60 s. This measurement gives the signal height for the As(V) species.
	15. Close solenoid valve SV3 and open solenoid valve SV1 to allow the contents of the reaction chamber to drain off.
	16. Close solenoid valves SV1 and SV2.
	17. Aspirate 2.5 mL of water and dispense to the reaction chamber. Repeat this step one more time so that the reaction chamber is flushed with 5 mL of water.
	18. Open solenoid valves SV1 and SV2 so that the contents are drained out.
	19. Close solenoid valves SV1 and SV2.
	
	As mentioned before, in the case of the portable field instrument the syringe pump is connected to a six-port distribution valve and one of the ports of the six-port distribution valve is connected to the common port three-way solenoid valve. While working in the sequential analysis mode, this valve should be energized before step 1, so that the citrate buffer is connected to the distribution valve and de-energized after step 4. The signal heights for As(III) and As(V) are then fed into the corresponding calibration curves to obtain the amount of the respective As species. The time required for one complete sequential analysis is about 6 min.
	
	

The air flow rate through the ozone generator was optimized at 60 mL/min, and the ozone produced at this flow rate was measured iodometrically to be 0.25% v/v. At lower flow rates the ozone concentration decreased, and at increased flow rates the amount of ozone produced was diluted. The ozone generator was operated at a duty cycle of about 30%. Increasing the duty cycle resulted in excessive heat generation, which reduced the ozone concentration.

 The chemiluminescence from the arsine-ozone reaction is not instantaneous and needs some finite amount of time to reach a maximum. Hence, the chemiluminescence signal intensity was optimized by varying the length of the black PTFE tube carrying arsine inside the chemiluminescence chamber. The chemiluminescence signal intensities are at a maximum when the PTFE tube has been withdrawn 4.1 cm from the tip of the glass tube, so that only 0.5 cm of it is inside the glass tube.

 With addition of sodium borohydride in the reaction chamber, about 50 mL of hydrogen is also evolved along with arsine, and this hydrogen can be used to purge the reaction chamber. However, when the reaction chamber was purged with air at a flow rate of 30 mL/min, the signal intensities reached their maximum. Increasing the flow rate further decreased the signal intensity by diluting the arsine. The same air purge also helps to drain out the contents of the reaction chamber quickly.

Reduction of As(V) to arsine was observed to be slower than the reduction of As(III). This difference in reduction rate resulted in a significant difference in chemiluminescence signal heights of the two species. To overcome this signal height difference, the solenoid valves SV2 and SV3 were kept closed for 30 s after the addition of sodium borohydride, and arsine gas was allowed to accumulate and then was released into the chemiluminescence chamber as a pulse. With an accumulation time > 30 s, the signal intensities for As(V) and As(III) became essentially equal.
	
	

	The results with As(III) standards in the concentration range 1–100 μg/L are shown in Figure 3. The inset figure shows the comparison of the blank readings with 1 and 2 μg/L of total arsenic. The blank readings show a chemiluminescence signal intensity of 0.067 ± 0.001 V, which resulted from the As present as an impurity in the sulfuric acid. As mentioned above, under strongly acidic conditions (pH ≥ 1), both As(III) and As(V) are very efficiently converted to arsine. With an accumulation time of 30 s, the As(III) and As(V) chemiluminescence signals are indistinguishable. Hence, either As(III) or As(V) standards can be used as calibrant for total As assays. The peak height and the peak area both produce a linear response in the range of interest (1–100 μg/L). For ease and simplicity, peak height has been plotted to obtain the following calibration curve:
	
	Peak height, V = 0.1923(±0.0015), 
	total As (μg/L) + 0.0675(±0.0016)  [1]
	

 = 0.9992                                   

	The corresponding calibration data for total arsenic measurement are shown in Figure 4. For a 2.5-mL sample, the limit of detection (LOD) (for S/N = 3) based on the standard deviation of the blank is estimated to be 0.03 μg/L.

	Both the laboratory instrument and the field instrument can measure As(III) and As(V) separately. Operating first at pH 4.5, the response is mostly from As(III) with a small response from As(V). Subsequent strong acid measurement of this sample results in a response that is predominantly from As(V) with a small response from residual unreacted As(III). The software then solves the relevant bivariate calibration equation to produce the final results. The actual response data from different mixtures of As(III) and As(V) are shown in Figure 5. The relevant bivariate equations (signal representing peak height) are as follows:
	
	As(III), μg/L = 9.74*signal 1,
	volts - 0.887*signal 2, volts      [2]
	
	As(V), μg/L = -1.35*signal 1,
	volts + 10.6*signal 2, volts          [3]

	It is well known that ozone produces chemiluminescence with many volatile gases. However, compared to the chemiluminescence from arsine, most of the volatile gases produce weak chemiluminescence. Idowu and colleagues (11) have previously shown that under the above test conditions this approach does not suffer from significant interferences from other common waterborne species, including bicarbonate, nitrate, sulfate, and, notably, sulfide. They also tested various tap water samples and other samples and carried out a blind intercomparison with the U. S. Geological Survey. Results from the gas-phase chemiluminescence technique were in very good agreement with those obtained with benchmark techniques, notably graphite furnace atomic absorption spectrometry (GFAAS) and liquid chromatography–ion chromatography–inductively coupled plasma–mass spectrometry (LC–IC–ICP-MS).

	Arsenic-contaminated drinking water is a serious issue, not only in developing countries but also in industrialized nations. The widespread use of hydraulic fracturing is only likely to exacerbate this problem. Globally about 140 million people are affected by arsenic poisoning. The need to develop better, cost-effective, field-deployable instruments for speciated arsenic analysis is a real societal issue in places that can least afford expensive instrumentation. We have described here the use of a fully automated, extremely sensitive instrument that can be assembled at modest cost and thus can potentially replace test kits that use toxic mercury and lead compounds. It will provide regulatory bodies with more elaborate data and a better understanding of arsenic levels to make better informed and accurate decisions.

	Visible luminescence from the arsine–ozone reaction is highly selective and sensitive, and has been known for a long time (12,13). If all the automated fluid handling were omitted and the instrument could still provide accurate quantification near the regulatory limit of 10 μg/L as has already been proven (14), one would have an even more affordable instrument for measuring arsenic in the field and could do so more rapidly and more accurately than possible with currently available technologies. Future work involves development of an instrument that can be used for industrial purposes based on continuous electrochemical reduction of arsenic to arsine, which would help monitor arsenic levels in large-scale continuous-flow operations.

	This work was supported by the National Science Foundation grant PFI: AIR-TT IIP-1414383. We thank Scott Evans of Lumion Laboratories, Inc., for market research.

	(1)   M.K. Sengupta, A. Mukherjee, M.A. Hossain, S. Ahmed, M.M. Rahman, D. Lodh, U.K. Chowdhury, B.K. Biswas, B. Nayak, B. Das, K. C. Saha, D. Chakraborti, S. C. Mukherjee, G. Chatterjee, S. Pati, R. N. Dutta, and Q. Quamuzzaman, 

304–310 (2000).
	(3)   J. Nriagu, P. Bhattacharya, A. Mukherjee, J. Bundschuh, R. Zevenhoven, and R. Loeppert (Eds.), 

Arsenic in Soil and Groundwater Environment

 (Elsevier, Amsterdam, 2007), pp. 3–60.
	(4)   United States Geological Survey. National Water Quality Assessment. 

http://water.usgs.gov/nawqa/trace/arsenic

.
	(5)   B.E. Fontenot, L.R. Hunt, Z.L. Hildenbrand, D.D. Carlton Jr., H. Oka, J.L. Walton, D. Hopkins, A. Osorio, B. Bjorndal, Q.H. Hu, and K.A. Schug, 

 10032–10040 (2013).
	(6)   J.A. Gomesa, D. Cocke, S. Varma, H. Morenoa, and E. Peterson, 

(14), 57–70 (2007).
	(7)   J. Feldmann, 

 263 (1879).
	(9)   A. Hussam, M. Alauddin, A.H. Khan, S.B. Rasul, and A.K. Munir, 

 3686–3688 (1999).
	(10) D.G. Kinniburgh, and W. Kosmus, 

 165–180 (2002).
	(11) A.D. Idowu, P.K. Dasgupta, Z. Genfa, and K. Toda, 

 7088–7097 (2006).
	(12) K. Fujiwara, Y. Watanabe, K. Fuwa, and J.D. Winefordner, 

 125–128 (1982).
	(13) M.E. Fraser, D.H. Stedman, and M.J. Henderson, 

 1200–1201 (1982).
	(14) M.K. Sengupta, Z.A. Hossain, S.I. Ohira, and P.K. Dasgupta, 

 are with the Department of Chemistry and Biochemistry at the University of Texas at Arlington in Arlington, Texas. 

 is with the University of Texas at Arlington Research Institute in Fort Worth, Texas. Please direct correspondence to: 

Articles

This cost-effective approach has a limit of detection well below 1µg As/L and a linear range that extends to >100 µg As/L.

A Fast, Accurate, Speciation-Capable, Automated, and Green Gas-Phase Chemiluminescence Approach for Analyzing Waterborne Arsenic

Here we highlight some of the opportunities associated with combining advanced sample preparation techniques with state-of-the-art chemical analysis techniques. This article considers the unique combination of selective pressurized liquid extraction (SPLE) with gas chromatography (GC) coupled with mass spectrometry (MS) and ultrahigh-pressure liquid chromatography coupled with tandem MS (UHPLC–MS-MS). We use this powerful combination to develop a novel analytical technique capable of measuring hormones and organic contaminants in whale earwax plugs. We explore the analytical challenges with such combinations and the advantages of focusing both on sample preparation as well as chemical analysis. 

	
	Sample preparation is a critical step in the analysis of organic compounds in solid and semisolid matrices and often involves multiple labor- and time-intensive steps, which in turn propagate uncertainty. Over the past decade, the automated extraction technique pressurized liquid extraction (PLE) has made significant advances to sample preparation as compared to classic techniques such as Soxhlet extraction. PLE is considered an exhaustive extraction technique suitable for a wide range of solid and semisolid matrices including tissues, soils, sediments, and particulate matter, as well as consumer products. To improve the extraction efficiency, PLE utilizes both high pressure (1500 psi) and high temperature (30–200 °C). Elevated temperatures serve to increase the solubility of the target analytes through accelerated extraction kinetics. Increasing the pressure helps ensure that the extraction solvent is below its boiling point. PLE typically uses between 20 and 100 mL of extraction solvent (depending on the size of the extraction cell) with extraction times of 15–30 min (depending on the static time and number of cycles). The reduction of extraction time and solvent are significant improvements compared to techniques like Soxhlet extraction and serve to increase analytical throughput. 
	
	PLE methods are optimized for specific analytes (typically with similar physical and chemical properties) through the assessment of solvent, temperature, extraction time, percent flush volume, number of cycles, and mass of matrix. During the extraction, target analytes and potential interference are extracted and collected together in a collection bottle. As a result, various cleanup strategies, such as packed chromatographic columns or gel-permeation chromatography, have been used to help remove potential interferences present in the extract. Complex matrices, such as biological tissues, often require multiple cleanup steps including the use of multiple packed chromatographic columns. These additional steps increase the time, training, space, and cost associated with the chemical analysis. Analytical trade-offs arise as matrix complexity and the number of target analytes increase.

	Recently, PLE techniques have integrated various cleanup strategies within the extraction cell and are often described as selective PLE (SPLE) or in-cell cleanup. SPLE is accomplished by layering commercially available adsorbents below the sample homogenate within the extraction cell. A wide range of commercially available adsorbents have been successfully incorporated within the extraction cell such as silica gel (including silica gel modified with sulfuric acid and potassium hydroxide), alumina, Florisil, carbon, diatomaceous earth, and C

. Subedi and colleagues (1) provided a detailed review of SPLE applications for organic contaminants.

	The incorporation of cleanup adsorbents within the extraction has been shown to eliminate some or all of the subsequent cleanup steps. The reduction of the number of analytical steps is advantageous on many levels including a reduction of training, space, and cost of consumables, as well as an increase in the overall analytical throughput and efficiency (that is, a reduction in analytical bottlenecks). These high-throughput SPLE methods also reduce or minimize the risk of sample contamination associated with sample preparation steps. Sample contamination is a major concern when measuring trace-level concentrations of organic contaminants in rare or irreplaceable samples.

	Marine mammals are often considered ecosystem sentinels because their survival is contingent on the health and functioning of marine ecosystems (2). Baleen whales (suborder 

) are unique sentinels because of their low trophic position and extensive geographic range, meaning that they are especially vulnerable to environmental perturbation. The bowhead whale (

), an Arctic species, is a sentinel in a habitat under rapid transition. As the extent and quality of sea ice decreases, this habitat is experiencing burgeoning impacts from oil and gas exploration, shipping activity, ocean noise, persistent contaminants, and commercial fishing-in addition to the ongoing ecological effects of climate change (3–5). Understanding and quantifying the exposure and effects of these stressors, such as persistent organic pollutants (POPs), on the bowhead whale is paramount to their conservation and management.

	Baleen whale earplugs are waxy structures that form annual or semiannual layers or lamina, and have classically been used in aging techniques (6,7). This waxy matrix consists of both a high lipid content (light layer) and keratinized epithelium cells (dark layer) (8). Recently, an SPLE technique was developed to measure a wide range of POPs in whale earwax (that is, cerumen [9]). One of the major advantages was the complete elimination of postextraction cleanup steps, which was accomplished by combining the necessary cleanup steps in the extraction step (9). In addition, enzyme-linked immunosorbent assay (ELISA) techniques have been developed to measure hormones in whale cerumen (10). It is important to note that each immunoassay kit is only capable of measuring a single hormone and that the analysis of multiple hormones would require significant sample mass, which may eliminate additional chemical measurements.

	POPs and hormone profiles were reconstructed for an individual whale with an estimated 6-month resolution, using the above mentioned methods to measure POPs and hormone concentrations in individual lamina (10). As a result, cerumen, and thereby earplugs, have the capability to record and archive chemical signals (both natural and anthropogenic). This technique is similar to chemical reconstruction techniques used in sediment or ice cores (11). This technique provides an unprecedented glimpse at a whale’s lifetime reproductive history, stress response, and contaminant exposure. This approach provides more accurate estimates of reproductive rate and age at sexual maturity than traditional methods and also yields baseline information regarding stress and contaminant exposure. One of the major issues associated with reconstructing chemical profiles using whale cerumen is the limited sample mass. Typically, the sample mass available for chemical analysis depends on the size of the earplug and, therefore, the age and species of the whale, but often ranges between 0.5 to 1.5 g/lamina. This limited sample mass reduces the overall number of chemicals or chemical classes that can be examined using a single earwax plug.

	The objective of this study is to expand on the SPLE method (9) to also include hormones, while preserving its ability to measure a wide range of POPs (Figure 1). Expanding the SPLE method would help maximize the number and type of analytes that can be reconstructed from a single whale earplug. Here, POPs are extracted using SPLE and extracts are analyzed using gas chromatography–mass spectrometry (GC–MS) (9). Hormones are eluted during a secondary extraction and those extracts are analyzed using ultrahigh-pressure liquid chromatography–tandem mass spectrometry (UHPLC–MS-MS). Liquid chromatography (LC) offers many advantages compared to traditional hormone analysis with immunoassay because smaller sample volumes can be used, multiple compounds can be measured from a single sample, and issues with cross reactivity are avoided (12). Hormones, such as testosterone, are typically more polar than the current list of POPs measured in whale earwax and offer a very unique set of analytical challenges. 


	
	This line of research has two main analytical challenges: First, developing an analytical technique capable of measuring part-per-billion concentrations of biologically relevant hormones in whale cerumen using UHPLC–MS-MS, and second, developing an SPLE technique that is capable of extracting contaminants as well as hormones from a single sample of whale cerumen without any additional cleanup steps.
	
	To optimize the SPLE technique, we need to maximize the extraction efficiency of the analytes from the matrix, in this case cerumen. To do so, we must simultaneously preserve the retention of our biological matrix compounds while extracting both organic contaminants and hormones. We believe we can extract both classes of compounds with the first extraction, but retain the hormones in the adsorbents along with the matrix itself. To do so, we must use adsorbents capable of retaining the lipid-rich matrix and hormones. Subsequently, the adsorbents should retain the bulk matrix while allowing the elution of the hormones with a secondary solvent extraction. A variety of solvents will be examined to produce the desired results.
	
	

	Unlabeled progesterone, estradiol, testosterone, cortisol, and labeled D

-cortisol with a purity of ≥98% were purchased from Cambridge Isotope Laboratories. All supplies were stored according to the manufacturer’s recommendations. Alumina (basic, activated, Brockmann I), and Florisil (60–100 mesh) were purchased from Sigma Aldrich. Silica gel (pore size 60 Å, 70–230 mesh) was purchased from BDH Chemicals. Concentrated formic acid and UHPLC-grade acetonitrile, ≥99.9% purity, were purchased from Fisher Scientific. For the UHPLC separations, a 100 mm × 2.1 mm, 1.7-µm 

 Acquity UPLC BEH C18 column and a 5 mm × 2.1 mm, 1.7-µm 

 UPLC BEH C18 VanGuard precolumn were purchased from Waters. Solvents and mobile phase were prepared using purified water (Thermo Barnstead Nanopure Diamond UV water purification, 18 Mâ¦).
	
	

SPLE for Organic Contaminants in Whale Earwax

	The SPLE technique capable of measuring contaminants (including pesticides, polychlorinated biphenyls, and polybrominated diphenyl ethers) in whale cerumen has been described previously (9). Briefly, aliquots of whale cerumen (∼0.25 g) were homogenized with anhydrous sodium sulfate and placed on precleaned adsorbents within the extraction cell (from top to bottom; 5 g of basic alumina, 15 g of silica gel, and 10 g of Florisil). All extractions were performed using an accelerated solvent extractor (ASE; ASE 350, Dionex [now part of Thermo Fisher Scientific]). Homogenates were extracted under the following extraction conditions with 1:1 dichloromethane–hexane, 100 °C, 1500 psi, 2 cycles (2 min each), and 150% rinse volume. Homogenates were spiked with isotopically labeled standards to correct for variability in analyte loss during sample preparation before SPLE. Extracts were spiked with a secondary set of isotopically labeled standards and concentrated to ~300 mL before GC–MS with electron ionization and GC–MS with electron capture negative ionization analysis.
	
	

	Hormone separation and analysis were performed using a Waters Acquity UPLC system and a Waters Xevo ESI/TQ-S electrospray ionization tandem MS system. Acetonitrile and water, with 0.1% formic acid, were selected along with a 100 mm × 2.1 mm, 1.7-µm 

 Acquity UPLC BEH C18 column with a 5-mm guard column with identical packing material and diameter. All samples were dried and reconstituted in 95:5 (v/v) water-acetonitrile with 0.1% formic acid following extraction. The column was initially equilibrated at 95% water (mobile-phase A) and 5% acetonitrile (mobile-phase B), for 30 min with a 0.5-mL/min flow rate. A quantitation method, using area, was established to monitor the following reactions (see Table I).

	Before analysis, the UHPLC system was flushed with 15 injections of water and 15 injections of acetonitrile at 0.7 mL/min with a 50% volume of 0.1% formic acid in water (mobile-phase A) and 0.1% formic acid in acetonitrile (mobile-phase B) to remove any waste from previous use. After it was cleaned, the column was loaded onto the instrument and conditioned at 95% A and 5% B for 30 min at 0.5 mL/min at 35 C. A target analyte calibration was performed with five points ranging from 2 ppb to 100 ppb using an isotopically labeled internal standard, then three blanks of 0.1% formic acid in water. The separation was programmed to begin at 5% B and increase to 40% B at 0.5 mL/min after 0.5 min and then gradually increase to 60% B for 4.5 min. The column was returned to initial conditions for 3.5 min before the separation was repeated. After the batch was complete, the column was flushed at initial conditions for 30 min and then flushed for 10 min in 95% B for storage.

SPLE Optimization for Hormones in Whale Earwax

	Typically, to optimize an SPLE method, a wide range of adsorbents, adsorbent masses, and combinations of adsorbents would be examined. However, because we are expanding on a previous SPLE method developed for the extraction and analysis of organic contaminants in whale earwax, the SPLE adsorbents have already been selected (adsorbents [basic alumina, silica gel, and Florisil]). At this juncture, we shift our focus to the hormones and acknowledge that hormones may be extracted from the whale earwax during the first extraction (extraction of contaminants using 1:1 dichloromethane–hexane) but retained on one or more of the adsorbents. In fact, our goal would be to extract both the contaminants and hormones during the first extraction step (dichloromethane–hexane) but retain the hormones for a secondary extraction on the adsorbents. In this ideal situation, the secondary extraction would elute the hormones from the adsorbents but preserve the retention of matrix compounds. The efficiency of extractions and adsorbents (retention of target analytes) were examined using target analyte percent recoveries and sample cleanliness (for example, effectiveness of adsorbents in retainment of interferences).

	The first step in optimizing the SPLE method for hormone analysis is to better understand the extraction of hormones from whale earwax using the prescribed SPLE adsorbents and solvent (dichloromethane–hexane; see below). This step is hindered such that specific hormones may be extracted from the earwax, but retained by one or more adsorbents. Hormones measured in the first extraction (dichloromethane–hexane) with high percent recoveries (>50%) would not be sufficiently retained in the extraction cell. Hormones with nondetects or very low recoveries (<10%) are assumed to be extracted, but retained by one or more adsorbent during the first extraction (dichloromethane–hexane). Past this point, a series of experiments must be designed to identify which hormones are retained by which adsorbents and which solvents or combinations of solvents are necessary to elute the hormones from the adsorbents.

	The extraction efficiency of multiple solvents and combinations of solvents were investigated as a potential second round extraction solvent (focusing on hormone elution). Dichloromethane, ethyl acetate, toluene, 1:1 dichloromethane–ethyl acetate, 2:1 ethyl acetate–dichloromethane, and cyclohexane were selected and examined based on literature information and polarity. Extractions were performed using a 100-mL cell containing 15 g of Florisil, 22.5 g of silica gel, and 7.5 g of basic alumina. All cells were conditioned (precleaned) using dichloromethane–hexane. Precleaning conditions were consistent throughout the experiments unless noted otherwise. Adsorbents were precleaned with a 1:1 dichloromethane–hexane at 100 C, 1500 psi, four cycles each with 2-min static times, a 60-s purge, and 50% flush volume. After precleaning, the cells were spiked with a solution of target analytes and left to equilibrate for 1 h at room temperature.

	The first extraction consisted of dichloromethane–hexane as previously described. The PLE conditions for each extraction were constant at 100 C, 1500 psi, two cycles each with 5-min static times, a 100-s purge, and 100% flush volume. 

	A following secondary extraction examined the extraction efficiency of different solvents (see above) and their ability to elute hormones retained on the adsorbents. In some specific cases, a third extraction was also examined and focused on very polar solvents including ethyl acetate and acetone. The second and third extractions used the PLE parameters described above and only varied by the solvent being examined. After the initial extraction of dichloromethane–hexane, all subsequent extractions were collected and analyzed separately. The samples were concentrated to ~1 mL using a Turbo Vap II evaporator (Biotage), then transferred to a GC vial and blown to dryness using compressed nitrogen with a fine blow-down peripheral. After drying, the sample was reconstituted in 475 mL of mobile phase-0.1% formic acid and 5% acetonitrile in water-then spiked with 25 mL of isotopically labeled internal standard before analysis using UHPLC–MS-MS.

	Target analyte affinity for individual adsorbents was examined with target analyte spike and recover experiments (

 = 1). Next, 10 g of each adsorbent were added to a 33-mL PLE cell. After conditioning (with dichloromethane–hexane; see above), each cell was spiked with a solution of target analytes as described above. Hormones were spiked on individual adsorbents packed into the extraction cell and were subsequently extracted using a range of extraction solvents. The extracts were then concentrated and reconstituted in the mobile phase as described above.

	Target analyte recoveries were also investigated with whale cerumen (0.25 g). A homogenate of multiple cerumen lamina created from a female bowhead whale (estimated at 65 years of age). A 0.25-g aliquot of the cerumen homogenate was homogenized with sodium sulfate (~10 g) and spiked in the extraction cell with target analyte. Cerumen samples, which provided a source of matrix interferences, were used to examine the matrix effects on 2:1 ethyl acetate–dichloromethane as the secondary extraction solvent. Extracts containing cerumen required a filter step to help remove precipitation, which appeared during blowdown. Filtering occurred after the extract was blown down to dryness and reconstituted in mobile phase (475 mL). A 13-mm Acrodisc CR filter with a 0.2-µm PTFE membrane (Pall Life Sciences) was precleaned with 500 mL of purified water. The samples were then drawn into a syringe and extruded through the filter into a clean GC vial.

	A third extraction was performed on two 66-mL cells, one with and one without earwax, both contained all three precleaned adsorbents. An extraction of ethyl acetate–dichloromethane (2:1) was performed on both cells. Once the extraction was complete the cells were allowed to cool to return to room temperature. The cell that did not contain any cerumen was extracted with ethyl acetate and subsequently with acetone. The cell cap of the cell containing wax was removed and the PLE extraction filter insertion tool was used to push the cellulose filter below the adsorbents upward. The layer of sodium sulfate and wax homogenate was scraped off carefully to leave behind the majority of basic alumina. After the homogenate was removed the sample was extracted with ethyl acetate followed by acetone.

	We began this novel method development with four hormones: cortisol, estradiol, progesterone, and testosterone. Hormone retention of each adsorbent was examined by placing 10 g of each adsorbent into a 33-mL extraction cell. The cells were conditioned with dichloromethane–hexane and then extracted with a range of different solvents. Analysis revealed that cortisol was well retained by all three adsorbents and preliminary results suggest that no solvent or combination was capable of eluting cortisol. Solvents and solvent combinations examined including 1:1 dichloromethane–ethyl acetate, 2:1 ethyl acetate–dichloromethane, ethyl acetate, toluene, ethyl acetate–toluene, and cyclohexane. The 2:1 ethyl acetate–dichloromethane solvent was selected over other solvents based on its ability to elute estradiol, progesterone, and testosterone. Toluene was capable of eluting select hormones, but required extensive blow-down time (3× as compared to 2:1 ethyl acetate–dichloromethane). Again, all solvent and solvent combinations were unable to elute cortisol off the adsorbents.

	Adsorbent experiments suggest that testosterone retention was dominated by silica gel. Spike and recovery experiments performed with all three adsorbents using 2:1 ethyl acetate–dichloromethane provided greater than 50% recovery. Percent recoveries were calculated by dividing the measured concentrations in the sample extract by expected concentrations and multiplying the quotient by 100%. A second experiment was performed in which cerumen (0.25 g cerumen homogenized with sodium sulfate) was loaded into a precleaned 66-mL cell and spiked with target analytes and left to equilibrate for 1 h. After it was extracted with 2:1 ethyl acetate–dichloromethane, the wax and sodium sulfate homogenate were extruded from the top of the cell and the three adsorbents were extracted with ethyl acetate. The ethyl acetate extraction provided a greater than 40% recovery of testosterone (Figure 2a).

	Adsorbent experiments suggest that estradiol retainment was dominated by Florisil. Spike and recovery experiments (without cerumen) performed with all three adsorbents using 2:1 ethyl acetate–dichloromethane provided greater than 80% recovery. Spike and recovery experiments (with cerumen) performed with all three adsorbents using 2:1 ethyl acetate–dichloromethane provided greater than 55% recovery. Again, a third extraction using ethyl acetate was performed after removing the sodium sulfate and wax homogenate. The ethyl acetate extraction provided a greater than 160% recovery of estradiol (Figure 2b). The higher estradiol recoveries were most likely caused by the native estradiol present in the female whale.

	Preliminary data suggest that hormones can be measured in whale earwax using an SPLE technique with a combination of extractions. Further parameters will be examined to improve the overall extraction efficiency, including the percent of flush volume, number of cycles, and static time. Future analysis will also include isotopically labeled target analytes to serve as surrogate standard. Surrogates will help correct for target analyte lose and variability in sample preparation.

	(1)   B. Subedi, L. Aguilar, E. Robinson, K. Hageman, E. Björklund, R. Sheesley, and S. Usenko, 

 119–132 (2015). 
	(2)   S.E. Moore, 

 534–540 (2008).
	(3)   C. Granier, U. Niemeier, J.H. Jungclaus, L. Emmons, P. Hess, J.-F. Lamarque, S. Walters, and G.P. Brasseur, 

(13), L13807 (2006).
	(4)   G.C. Hays, A.J. Richardson, and C. Robinson, T

 337–344 (2005).
	(5)   D.K. Perovich and J.A. Richter-Menge, 

 417–441 (2009).
	(6)   C.M. Gabriele, C. Lockyer, J.M. Straley, C.M. Jurasz, and H. Kato, 

 443–450 (2010).
	(7)   A. Jonsgard in 

 H.T. Andersen, Ed. (Academic Press, New York, 1969), pp. 1–30.
	(8)   P.E. Purves, 

 293–302 (1955).
	(9)   E.M. Robinson, S.J. Trumble, B. Subedi, R. Sanders, and S. Usenko, 

 14–20 (2013).
	(10) S.J. Trumble, E.M. Robinson, M. Berman-Kowalewski, C.W. Potter, and S. Usenko, 

 16922–16926 (2013).
	(11) S. Usenko, D.H. Landers, P.G. Appleby, and S.L. Simonich, 

 7235–7241 (2007).
	(12) C.J. Hogg, E.R. Vickers, and T.L. Rogers, 

J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 

 is with the Department of Chemistry and Biochemistry and the Department of Environmental Science at Baylor University in Waco, Texas. 

 is with the Department of Chemistry and Biochemistry at Baylor University. 

 is with the Department of Biology at Baylor University. 

is with the Department of Environmental Science at Baylor University. Direct correspondence to: 

This work addresses two challenges: developing a technique capable of measuring ppb levels of hormones, and developing an SPLE technique capable of extracting contaminants and hormones from a single sample without additional cleanup steps.

GC–MS and UHPLC–MS-MS Analysis of Organic Contaminants and Hormones in Whale Earwax Using Selective Pressurized Liquid Extraction

Single-particle inductively coupled plasma–mass spectrometry (SP-ICP-MS) is an exciting new technique for detecting and characterizing metal nanoparticles at very low concentrations. It is fast and can provide significantly more information than other traditional techniques, including particle number concentration, particle size, and size distribution, in addition to the concentration of dissolved metals in solution. The added benefit of using ICP-MS is that it can distinguish between particles of different elemental compositions. The study will investigate the use of SP-ICP-MS to track the release of engineered nanoparticles (ENPs) into the environment and to better understand their fate and behavior specifically in drinking, surface, and wastewater samples. 

	
	The unique properties of engineered nanoparticles (ENPs) have created intense awareness in their environmental behavior. Because of the increased use of nanotechnology in consumer products, industrial applications, and healthcare technology, nanoparticles are more likely to enter the environment. For this reason, it is not only important to know the type, size, and distribution of nanoparticles in soils, potable waters, and wastewaters, but it is also crucial to understand their impact on the growing mechanism of crops used for human consumption. Therefore, to ensure the future development of nanotechnology products, there is clearly a need to evaluate the risks posed by these ENPs, which will require proper tools to fully understand their toxicological impact on human health. Current approaches to assess exposure levels include predictions based on computer modeling, together with direct measurement techniques. Predictions through modeling are based on knowledge of how they are emitted into the environment and by their behavior in the samples being studied. Although the life cycles of ENPs are now starting to be understood, very little is known about their environmental behavior. Prediction through life-cycle assessment modeling requires validation through measurement at environmentally significant concentrations. For ENPs that are being released into the environment, extremely sensitive methods are required to ensure that direct observations are representative in time and space. ENPs differ from most conventional ‘‘dissolved’’ chemicals in terms of their heterogeneous distributions in size, shape, surface charge, composition, and degree of dispersion. For this reason, it is not only important to determine their concentrations, but also these other important metrics, particularly when they are discharged and interact with their real-world surroundings.
	
	

Impact of Nanoparticles Released into the Environment


	
	When nanoparticles enter the environment, they can undergo a number of potential transformations that depend not only on the properties of the nanoparticles but also on the medium they are being released into. These changes typically involve chemical and physical processes, but they can also involve biodegradation of surface coatings used to stabilize many nanomaterials. For example, the toxicity impact of nanomaterials on algae involves adsorption onto cell surfaces, which has the potential to disrupt transportation through the cell membrane. Additionally, larger organisms can directly ingest nanoparticles, and then enter the food chain when they are consumed by other aquatic and terrestrial forms of life. These processes are further complicated by aggregation of nanoparticles with other natural minerals and natural colloids, which will dramatically change their fate and potential toxicity in the environment.
	
	Nanomaterials from domestic, medical, and industrial sources may also undergo significant changes during wastewater cleaning processes, such as the conversion of silver nanoparticles to silver sulfide in treatment plants. In addition, aggregation of the nanomaterials with other minerals and organic matter in the wastewater often results in the nanomaterial combining with other solids in the effluent, rather than remaining as a dispersed nanoparticle suspension. So, apart from atmospheric deposition or accidental spillage of nanomaterials directly into rivers, lakes, and the surrounding land, wastewater treatment remains by far the largest source of nanomaterial contamination either from the runoff into groundwater sources and drinking water supplies or from the raw sewage sludge that is often spread onto the soil as a fertilizer (1).
	
	
	


	
	Therefore, risk assessment for nanomaterial released into the environment is still evolving, and reliable measurements of environmentally significant concentrations remain challenging. Predicted environmental concentrations based on current usage are low, but they are expected to increase as the use of nanomaterials increases. At this early stage, comparisons of estimated exposure data with known toxicity data indicate that the predicted environmental concentrations are orders of magnitude below those known to have environmental effects on living plant and animal biota. As more toxicity data are generated under environmentally relevant conditions, risk assessments for nanomaterials will improve to produce accurate assessments that ensure environmental safety.
	
	Many analytical techniques applied in materials science and other scientific disciplines could also be applied to ENP analysis. However, environmental and biological studies may require that methods be adapted to work at extremely low concentrations in complex matrices. The most pressing research needs are the development of techniques that introduce minimal artifacts to optimize sensitivity, specificity, and throughput, as well as differentiating between naturally occurring particles and manufactured nanoparticles, together with a better understanding of their dissolution properties.


	
	The measurement and characterization of nanoparticles is therefore critical to all aspects of nanotechnology. In the field of environmental health, it has become clear that complete characterization of nanomaterials is important for interpreting the results of toxicological and human health studies. Metal-containing ENPs are a particularly significant class because their use in consumer products and industrial applications makes them the fastest growing category of nanoparticles (2).
	
	Although many analytical techniques are available for nanometrology, only some can be successfully applied to environmental and health studies. Methods for assessing particle size distributions include electron microscopy, chromatography, laser light scattering, ultrafiltration, and field flow fractionation. However, the lack of specificity of these techniques is problematic for complex environmental matrices that may contain natural nanoparticles having polydisperse size distributions and heterogeneous compositions. For this reason, extremely sensitive detection techniques are needed if specific information about the elemental composition and concentration of the nanoparticles is required. Unfortunately, difficulties can also arise with some detection techniques because of a lack of sensitivity for characterizing and quantifying particles at environmentally relevant concentrations.
	
	


	
	One technique that is proving invaluable for detecting and sizing metallic nanoparticles is single-particle inductively coupled plasma-mass spectrometry (SP-ICP-MS) (3). Its elemental specificity, sizing resolution, and unmatched sensitivity make it extremely applicable for the characterization of ENPs containing elements such as Ag and Au and compounds such as TiO

that have been integrated into larger products such as consumer goods, foods, pharmaceuticals, and personal care products.
	
	Much of the early work has focused on the use of ICP-MS with particle separation techniques, such as field flow fractionation and chromatography (4). However, more recently, SP-ICP-MS has shown a great deal of promise in several applications areas, including the determination of concentrations of silver nanoparticles in complex samples (5). This technique is suited to differentiate between the analyte in solution and existing as a nanoparticle without any prior separation techniques, simplifying nanoparticle analysis while eliminating complex sample preparation steps (6). This ability allows SP-ICP-MS to provide information on the size and size distribution of many varied and different metal-based nanoparticles, as well as the dissolved concentration of the analytes under study. Let’s take a closer look at the fundamental principle of SP-ICP-MS.

	SP-ICP-MS is based on the measurement of the signal intensity produced by a single particle. Nanoparticle suspensions are sufficiently diluted to minimize the chances that more than one particle reaches the plasma at a time. The particle is atomized and ionized, producing a signal of relatively high intensity, which is measured as a pulse. This process is exemplified in Figure 1, which shows particles (P) and dissolved analyte in the sample aerosol entering the plasma and being ionized. The ions then pass through the interface region into the ion optics where they are eventually separated in the mass spectrometer. In this manner, particles are detected as individual pulses, whereas the dissolved analyte contributes to a continuous background signal. The frequency of pulses (events) provides the particle number concentration, whereas the intensity of each pulse is proportional to the mass of the nanoparticle. Because of the short transient nature of the pulse, very short integration times are necessary to maximize the detection of individual particles as pulses of ions after they are ionized by the plasma.

	This process is exemplified in Figure 2, which shows both metal nanoparticles and metal ions in solution being ionized by an ICP-MS system. The signal from the dissolved ions is represented by the continuous signal below the dashed line, while the ionized pulses of nanoparticles are represented by the individual spikes. For this approach to work effectively, the speed of data acquisition and the response time of the detector must be fast enough to capture the time-resolved nanoparticle pulses, which are typically 300–400 µs (3). If the electronics are not fast enough, multiple or many pulses can easily pass through in a single integration window leading to inaccurate particle counting and sizing. For this application, the ICP-MS should be capable of using dwell times shorter than the particle transient signal time, thus avoiding false signals generated from clusters of particles. In practice, for an instrument that is optimized for nanoparticle characterization, this means using dwell times of 100 µs or less and a settling time of zero, so that the pulse can be fully characterized and precisely integrated using a peak area integration algorithm (7).

	It is important to emphasize that ICP-MS is a mass-based technique, so in SP-ICP-MS, the particle size is determined by relating the pulse intensity to an elemental mass. With traditional ICP-MS analysis, the first step in this process is to create a calibration curve using dissolved standards. This curve connects the signal intensity from the instrument to the concentration of the analyte entering the plasma. The next step is to relate the concentration of the dissolved analyte to a total analyte mass that enters the plasma during each reading. This relationship between analyte concentration and the mass observed per event is called the 

, which is highly dependent on the transport efficiency of the sampling process and the instrument ion optics. This transport efficiency must be calculated for each instrument and under the given run conditions for the mass flux to be accurate. In this way, the resulting calibration curve relates signal intensity (counts/event) to a total mass transported into the plasma per event. So by using well-understood SP-ICP-MS principles, the intensity of each individual pulse (counts/event) can then be transformed using the mass flux calibration curve to determine the particle mass, which is easily converted to particle diameter, by knowing the density and assuming that the geometry (shape) of the particle is spherical (8,9). This is exemplified in Figure 3, which shows a signal of multiple gold nanoparticles over time (Figure 3a), with an individual pulse on the left-hand-side at the bottom (Figure 3b) and the dissolved ionic calibration curve on the top (Figure 3c). The top right-hand-side of Figure 3d shows the mass of the particle (in micrograms) per dwell time after nebulizer efficiency and flow rate have been applied. The mass of the nanoparticle is then converted to a diameter, based on the mass fraction and the density, based on the shape of the particle (Figure 3e) (10). Note: If the nanoparticle is rod-shaped (for example) a different calculation is used.

	Now, let’s take a look at the real-world applicability of SP-ICP-MS by evaluating its capability and suitability for characterizing nanoparticles in environmental samples.

Real-World Applicability of SP-ICP-MS for Water Samples

	Silver (Ag) nanoparticles are among the most commonly used nanoparticles in consumer products, such as fabrics, deodorants, and detergents because of their antimicrobial properties. Therefore, it is expected that Ag ENPs will find their way into the environment, necessitating ways to accurately and rapidly detect and characterize them in a variety of environmental matrices. Work has already been performed demonstrating the ability to successfully characterize Ag ENPs and their dissolution characteristics in a variety of drinking water (11), surface water (12), and wastewater samples (13). Let’s take a closer look at these three very different application requirements for SP-ICP-MS. Note: All the data presented in the following section were generated on a NexION 350D ICP-MS system using the Syngistix Nano Application Software Module (PerkinElmer Inc.) (14). 

	SP-ICP-MS is an ideal tool for assessing the efficiency of drinking water treatment systems in removing silver, gold, and titanium dioxide (TiO

) nanoparticles without using any other analytical technique. To evaluate the effectiveness of the water treatment process for these types of nanoparticles, Donovan and colleagues (11) collected water samples both pre- and post-treatment at three water treatment plants in Missouri. None of the six waters contained measurable amounts of Ag or Au, either as particles or dissolved species. However, all source water samples contained TiO

, as shown in Table I. Plants 1 and 2 effectively removed both dissolved Ti and TiO

 particles, as evidenced by the lower amounts present in the post-treatment waters than the pretreatment samples. The results from Plant 3 differed from the first two plants in that Ti-containing particles could still be detected after treatment, although significantly less than the pretreatment sample. However, all the dissolved Ti was removed. 

	To check the accuracy of this method, spike recovery tests were performed for all metals, both as dissolved (Diss.) and particle (Part.) concentrations. Three samples were spiked with 2 μg/L of the dissolved metals and 100-nm nanoparticles at a concentration equal to 1 × 10

 particles/mL. Table II shows the results of these spike recovery studies, which indicate accurate recoveries, except for the concentration of TiO

 nanoparticles, which recovered below 10% for each sample. Low recovery for TiO

 particles was most likely because of aggregation in the standards and in the water matrix. Without the addition of a surfactant, these uncapped TiO

 particles tended to aggregate, which resulted in particle loss between dilutions and in the water matrix where they had an opportunity to react and form new species or aggregate further and fall out of solution. The highest recovery obtained for these particles in ultra-pure water and in a surface water matrix were 24% and 9%, respectively.

	For this work, Hadioui and colleagues (12) investigated the efficiency of SP-ICP-MS for the detection and characterization of metal nanoparticles in fresh surface waters where they can be involved in various physicochemical processes as shown in Figure 4. In this example, dissolved silver (A), including released free Ag ions (C, D) and soluble complexes (E) can easily and instantly be measured by SP-ICP-MS. These dissolved species could also be determined by ultrafiltration followed by total metal quantification using ICP-MS or atomic absorption spectroscopy (AAS), but this procedure is time consuming since it requires the pre-equilibration of the membrane for at least three cycles of centrifugation, generally 20 min each (15). Aggregates (B) and remaining stable silver nanoparticles (G) can be counted and measured by other commonly used techniques such as nanoparticle tracking analysis (NTA), dynamic light scattering (DLS), and transmission electron microscopy (TEM), but SP-ICP-MS is the only method that can distinguish between nano-Ag (nAg) and other metal-based colloids in surface waters.

	The surface water was sampled from a river in Montreal, Canada, and filtered with 0.2-μm filter paper before spiking with silver nanoparticles. Nano-Ag suspensions were added to water samples with concentrations ranging from 2.5 to 33.1 μg Ag/L and left to equilibrate under continuous and gentle shaking. Commercially available suspensions of gold and silver nanoparticles were used in this work. A National Institute of Standards and Technology (NIST) reference material (RM 8013) consisting of a suspension of gold nanoparticles (60 nm nominal diameter, 50 mg/L total mass concentration, and stabilized in a citrate buffer) was used to determine the nebulization efficiency, while two types of silver nanoparticle suspensions were used for the dissolution studies-a citrate-coated version (40 nm and 80 nm nominal diameter) and a noncoated version (80 nm nominal diameter). In addition, 60-nm nominal diameter Ag nanoparticles were used for quality control (QC) purposes.

	An interactive view of data acquisition of intensity as a function of time for the 60-nm nominal diameter silver nanoparticles is shown in Figure 5. In this example, the concentration of the analyte was 208 ng/L and was converted to volume and then into size knowing the density and the geometry of the particle using the Syngistix Nano Application Module, with no further need for manual data processing.

	It is also important to emphasize that even after filtration of surface waters at 0.2 μm, the NTA method, which is commonly used for this type of analysis, showed the presence of quite significant amounts of colloidal particles with an average diameter of ~110 nm. Therefore, the addition of metal nanoparticles to this complex matrix would make their detection and characterization very difficult, if not impossible, with the commonly used techniques previously mentioned, such as DLS and TEM. Furthermore, even the determination of the dissolved fraction that is usually performed by ultrafiltration may be inadequate because silver ions may adsorb on the surface of the colloids and, therefore, be retained by the filtration membrane. Consequently, the proportion of dissolved metal will be underestimated. SP-ICP-MS measurements were found to be more effective and to have fewer limitations than other techniques. Indeed, the presence of other insoluble particles did not interfere with the analysis of silver nanoparticles, as the signal of Ag is recorded independently of the other constituent elements of the colloids.

	The same group of researchers found that the SP-ICP-MS technique allowed the effective and selective measurement of changing particle size, aggregation, and dissolution over time at low concentrations. In fact, they concluded that SP-ICP-MS is probably the only suitable technique that can provide such information on the fate of metal nanoparticles at very low concentrations typically found in environmental waters (15).
	
	

	Another common, more complex matrix that must be evaluated for the determination of Ag ENPs is wastewater samples from wastewater treatment plants. The complexity and variety of the wastewater matrices can make the analysis of ENPs extremely challenging. Figure 6 is a representation of a wastewater treatment plant that shows the possible pathways that could introduce nanomaterials into the environment and the potential impact they might have on the surrounding land and water supplies (1).

	A recently published study by Azodi and colleagues (13) evaluated the ability of SP-ICP-MS to characterize Ag ENPs in three common wastewater matrices: mixed liquor, effluent, and alginate solutions. The wastewater samples were collected from a wastewater treatment plant near Montreal, Quebec, Canada. The effluent wastewater was collected after the secondary settling tank, while the mixed liquor was collected after the secondary aeration tank. The effluent wastewater is the final treated wastewater that is discharged to the river from this treatment plant, while the mixed liquor is the wastewater that leaves the aeration tank after biological treatment to settle the suspended solids in the settling tank. As a result, the mixed liquor has much higher levels of suspended solids and a relatively higher dissolved carbon content compared to the effluent wastewater.

	Alginate, more commonly known as alginic acid, is an anionic polysaccharide distributed widely in the cell walls of brown algae, which is present at parts-per-million levels in wastewaters and comprises the dissolved organic carbon fraction of wastewaters. The alginate solution was used as a known control-surrogate for comparison with the wastewater samples. All solutions were prepared using the alginic acid sodium salt from brown algae (at 6 ppm) in deionized water by shaking end-over-end for 1 h. Ag ENPs capped with polyvinylpyrrolidone (PVP) with a mean diameter of 67.8 ± 7.6 nm (as determined with TEM) were spiked into 10 mL of all samples at a concentration of 10 ppb (5 million particles/mL). The samples were then diluted 10–1000× with deionized water and sonicated for 5 min before analysis. All samples were prepared in triplicate. To determine the accuracy of SP-ICP-MS, the Ag ENPs were added to deionized water at a concentration of 0.1 ppb (50,000 particles/mL). The SP-ICP-MS measurement determined the mean size of the Ag ENPs to be 63.2 nm (which agrees with the TEM measurements), and the concentration to be 53,758 particles/mL, thus validating the accuracy of the measurements.

	Next, the Ag ENPs were measured in a 6 ppm alginate solution. Once the accuracy of the technique was ensured, it was followed by the measurement of the effluent wastewater and mixed liquor samples. First, the total Ag concentration was measured in both wastewater samples and was found to range from 25 to 40 ppt, levels that should not inhibit the determination of Ag ENPs. Figure 7a shows the Ag particle size distribution for 0.1 ppb (50,000 particles/mL) in the alginate sample, which corresponds to a mean particle size of 66.1 nm, with a concentration of 52,302 particles/mL. The agreement between the measured and TEM-determined particle sizes indicates that the alginate matrix does not affect the measurement accuracy as exemplified in Figures 7b and 7c, which show the measured particle size distributions for the effluent and mixed liquor respectively. Table III shows the measured particle sizes of both samples that agree with the certificate value, together with particle concentrations, which are close to the calculated value, indicating that neither of the wastewater matrices impact the measurement and strongly suggesting that Ag ENPs can be successfully measured in wastewater samples.

	The detection limits for both the Ag particle size and concentrations in the wastewaters were also determined. For determining the particle size detection limits, the diluted samples were analyzed without any Ag ENPs being added. The detection limit was determined by running the unspiked wastewater matrices and observing the particle size, which corresponded to the smallest recorded peak. For the effluent, the detection limit was about 18 nm, while for the mixed liquor, it was about 12 nm. To determine the lowest concentration of Ag ENPs that could be detected, Ag ENPs were spiked into deionized water and diluted multiple times (Note: Since the wastewater matrices were shown not to affect the results, the detection limit was only measured in deionized water). The measured particle concentration was then recorded for each concentration. The particle concentration detection limit was determined to be the spiked particle concentration where the measured concentration did not change when the sample was diluted. In this work, the Ag ENP particle concentration detection limit was determined to be 25 ppt (12,500 particles/mL).

	SP-ICP-MS demonstrates very exciting potential for the characterization of nanoparticles in many varied types of environmental samples. It is a very effective tool to assess the efficiency of drinking water treatment systems in removing certain nanoparticles, without the need for any additional techniques. It has also allowed the effective and selective measurement of changing particle size, aggregation, and dissolution over time at low concentrations in natural surface waters. Additionally, the technique has shown the versatility to measure and analyze nanoparticles in various matrices found in a wastewater treatment plant. There is no question that compared to other traditional analytical methods, SP-ICP-MS offers unique capabilities in tracking the fate and behavior of metal nanoparticles at extremely low levels in the environment. Although this study has focused on the effectiveness of the technique for the characterization of nanoparticles likely to be present in environmental water samples, it is also applicable to other types of metal and metal oxide nanoparticles in a variety of complex environmentally significant matrices including plant material (16) and biological tissue (17).

	(1)   G.E. Batley, J.K. Kirby, and M.J. McLaughlin, 


	(2)   The Project on Emerging Nanotechnologies, “An Inventory of Nanotechnology-based Consumer Products Introduced on the Market,” available at: 

 1143 (2012).
	(4)   D.M. Mitrano, A. Barber, A. Bednar, P. Westerhoff, C.P. Higgins, and J.F. Ranville, 

1131–1142 (2012).
	(5)   D.M. Mitrano, E.K. Leshner, A. Bednar, J. Monserud, C.P. Higgins, and J.F. Ranville, 

(1), 115–121 (2012).
	(6)   F. Laborda et al., 

(7), 1362–1371 (2011).
	(7)   A. Hineman and C. Stephan, 

1252–1257 (2014).
	(8)   A .C. Degueldre and P.Y. Favarger, 

 137–142 (2003).
	(9)   C. Degueldre and P.Y. Favarger, 

Colloids and Surfaces A: Physicochem. Eng. Aspects,

 137–142 (2003).
	(10) H.E. Pace, J. Rogers, C. Jarolimek, V.A. Coleman, C.P. Higgins, and J.F. Ranville, 

(24), 9361–9369 (2011).
	(11) A.R. Donovan, H. Shi, C. Adams, and C. Stephan, PerkinElmer Application Note, “Rapid Analysis of Silver, Gold, and Titanium Dioxide Nanoparticles in Drinking Water by Single Particle ICP-MS” (2015).
	(12) M. Hadioui, K. Wilkinson, and C. Stephan, PerkinElmer Application Note, “Assessing the Fate of Silver Nanoparticles in Surface Waters Using Single Particle ICP-MS” (2014).
	(13) M. Azodi, S. Ghoshal, and C. Stephan, PerkinElmer Application Note, “Measurement and Analysis of Silver Nanoparticles in Wastewaters with Single Particle ICP-MS” (2015).
	(14) PerkinElmer Application Study, “Single Particle ICP-MS Syngistix Nano Application Module,” 

http://www.perkinelmer.com/catalog/product/id/N8140309

.
	(15) M. Hadioui, S. Leclerc, and K.J. Wilkinson, 

 15–19 (2013).
	(16) Y. Dan, W. Zhang, X. Ma, H. Shi, and C. Stephan, PerkinElmer Application Note, “Gold Nanoparticle Uptake of Tomato Plants Characterized by Single Particle ICP-MS” (2015).
	(17) E.P. Gray et. al., 

 is the manager of global applications of nanotechnology at PerkinElmer, Inc., in Woodbridge, Ontario, Canada. 

 is the principal of his own freelance writing and scientific consulting company, Scientific Solutions, based in Gaithersburg, Maryland. Direct correspondence to: 
	

SP-ICP-MS demonstrates excellent potential for characterizing nanoparticles in varied types of environmental samples.

The Benefits of Single-Particle ICP-MS to Better Understand the Fate and Behavior of Engineered Nanoparticles in Environmental Water Samples

Gas chromatography (GC), inductively coupled plasma–mass spectrometry (ICP-MS), ICPâoptical emission spectrometry (OES), and other bulk analysis methods are applied to groundwater in proximity to unconventional oil and natural gas extraction activities. 

	
	The United States has experienced a dramatic shift in economic influence over the past 10 years with the widespread engineering advances that have allowed unconventional oil and gas (UOG) extraction to become more efficient and cost-effective. Small rural towns have become industry hubs overnight as a result of the hydrocarbons trapped beneath the ground. Educators have increased the number of engineering and technical programs available to students in an effort to meet the demand for a qualified workforce. Stories of multigeneration ranchers becoming millionaires overnight through the leasing of their land and mineral rights puts the television show 

in a more current light.
	
	Regulations related to UOG activity are currently left to each state, which creates disparities in environmental testing and monitoring across the United States. For example, Colorado and Illinois require baseline groundwater testing before drilling commences, while Pennsylvania only suggests baseline measurements in the event that a dispute arises after drilling. The list of organic compounds Colorado has chosen to monitor through chromatographic methods are total petroleum hydrocarbons, benzene, toluene, ethyl benzene, and xylenes (BTEX), polycyclic aromatic hydrocarbons (PAH) plus benzo[

]pyrene, and dissolved gases. These analytes have been the focus of analyses for years. Their presence in water is hypothesized to indicate an adverse environmental interaction with hydrocarbon extraction operations. However, their sources can still be convoluted and may not be wholly specific to UOG.

	Within the past decade, a mix of analytical methods has been developed or applied to establish an understanding of the impact, if any, that UOG activity is having on groundwater in the vicinity. This article discusses chromatographic methods applied for particular organic compounds and considerations to assist in method development. A section is also dedicated to spectroscopic methods for detection and quantification of metals and ions in water samples that are relevant to UOG activity. Collections from our groundwater research are highlighted in each section to demonstrate the application. 
	
	


	
	The ideal situation for creating a method would be to work with a system of 

(1) or compounds that are expected to be present and are positively identified. Current researchers must also have the capabilities to work with 

or the ability to determine unexpected identifiable compounds, and 

, compounds unexpected and without standards, Chemical Abstracts Service (CAS) numbers, or absent in databases. For example, these may include proprietary polymers or surfactants developed primarily for the UOG field. A hesitation that may be encountered are 

, which are expected compounds not detected. The internal debates are made up of how confident the compound is to be a “known;” is the concentration too low to detect in the given sample, or is the method inadequate? In the discussion to follow, there are very few “knowns” to be expected when monitoring groundwater possibly impacted by UOG activity.
	
	


	
	Gas chromatography (GC) methods have been at the forefront for analysis of organic compounds in groundwater and UOG wastewater (UOGWW) (2). While there is the potential for nonvolatile organic additives such as surfactants to be present, the majority of hydraulic fracturing additives or shale formation compounds of health or environmental concern are GC amenable. In a 2011 Congressional report, 24 organic hydraulic fracturing additives are listed as “Chemical Components of Concern,” of which 23 are GC amenable without the need for derivatization. Some of these include BTEX, diesel, and naphthalene, which have been suggested for baseline measurements by various states.
	
	Numerous Environmental Protection Agency (EPA) and state regulatory methods have been established using GC for these and a multitude of other compounds of concern over the past 50 years. While a mix of regulatory methods can be found that include a subset of these compounds, the lack of a single dedicated standard approach to effectively extract and separate a probable list of compounds in groundwater or UOGWW is a complicating factor that has slowed research. Most officially standardized versions of these methods are less capable of the throughput needed to prepare and analyze a large number of samples in a limited timeframe.

	The earliest efforts to assess the impact of UOG activity on groundwater was through the measurement of dissolved gases, specifically methane, ethane, propane, butane, and pentane (C

, respectively) in groundwater from regions within close proximity of UOG drilling sites (3). Methane is the most abundant component of natural gas extracted for energy purposes, with ethane and propane comprising the majority of the remaining small fraction. The hypothesis is that if there is a failure in the integrity of the protective casing of the UOG well (4,5) or if induced fractures in the shale create interconnectivity with the overlying aquifer, the natural gas would be the most abundant and mobile species to detect in groundwater.

	Two types of methane can be measured in groundwater (6). The most common type found in shallow groundwater is biogenic methane, a by-product of bacterial metabolism. Thermogenic methane is the other type, the primary target of UOG recovery. This methane gas is formed by the presence of decomposing organic matter under high temperatures and pressures over a long period of time (that is, from deep geological formations). Because of the different implications for each type of natural gas, methane measured in shallow groundwater must include further investigations to distinguish between biogenic or thermogenic origins.  

	The origin of the measured methane can be determined either through isotopic abundances of carbon-13 (

H), or the methane to ethane and propane ratio (7). A ratio of methane to higher chain hydrocarbons of less than approximately 100 suggests thermogenic gas (3). Both of these approaches have even been found to not only identify thermogenic methane, but also distinguish between different natural gases produced in different geological formations (3,8,9).

	GC separations coupled with flame ionization detection (FID) are most commonly used for the analysis of these light hydrocarbons. Groups measuring these light hydrocarbons do so in a targeted manner, meaning they are tailored specifically for C

 gases and little else. These methods are quite sensitive and selective, but ultimately lack the ability to detect a wide suite of unknown compounds. Sample introduction is performed through either purge-and-trap techniques (10), where the water is purged with an inert gas and volatiles are trapped on a selective sorbent, or using headspace analysis (11), where the sample is heated and agitated to liberate the gas to an open headspace in the vial, which is then sampled. Column selection for this analysis leans toward the use of porous layered open tubular (PLOT) columns, primarily those with a divinylbenzene phase (10,12). As alluded to earlier, PLOT phases possess a great affinity for C

 hydrocarbons, but that affinity is further extrapolated to the C

 linear and branched alkanes and aromatics, which are also present in natural gas, making analysis of these larger hydrocarbons inefficient because of long retention and excessive band broadening. An example of this affinity is demonstrated in Figure 1 with a natural gas standard separated on a PLOT divinylbenzene column and a 5% diphenyl capillary column.

	Intricate sample collection further complicates this specific analysis, which typically requires additional measurements, such as isotopic analysis, for results that can enable sourcing of the natural gas in the water (that is, a comparison of the natural gas isotope signature in the water with that from the targeted shale or other sources). Groundwater samples are typically collected from volunteers’ water wells, in which the withdraw rate and consistency will be variable across the population. Agitation, along with the pressure differential on the water once it reaches the surface, can cause the water to degas and skew dissolved gas measurements to below their actual values. For better control during sampling, it is recommended to use a nongas permeable tube with a valve connected to the water well head that flows at a constant rate into an evacuated bladder (for example, an IsoFlask sampling bladder [Isotech Laboratories]). This sampling bladder should be preloaded with a chosen biocide to reduced degradation of the gas by bacteria, on top of being stored at 4 °C for a 14-day maximum holding time before analysis (13).   

	As a complementary approach, our group has also demonstrated the capabilities of a new spectroscopic detector, the VGA-100 vacuum ultraviolet (VUV) detector (14) (VUV Analytics), which measures gas phase absorption in the VUV and ultraviolet wavelength regions (120–240 nm) to monitor dissolved gases in water. This universal detector offers qualitative gas-phase VUV spectra to accompany the quantitative capabilities of the C

 if interested (15). While this work separated C

 hydrocarbons of three water samples from the Barnett Shale with the HP-PLOT Q column (30 m × 0.32 mm, 20-µm 

), the deconvolution capabilities of the acquired spectra could allow the compounds to be quantified in the void volume of capillary columns. More work is needed to interface this detector with the various sampling protocols that exist for measuring natural gas in water to demonstrate its unique qualitative and quantitative capabilities for routine analysis. The blue chromatogram of Figure 1 is the previously discussed GC separation of a natural gas standard with the PLOT divinylbenzene column and VUV detection. 

	In the vast majority of UOG reservoirs, hydraulic fracturing is used to stimulate the formation. The fluids used to open fissures in low permeability shale formations include water, sand, and a small percentage of chemicals. These chemicals are a mixture of acids, bases, salts, organic compounds, and inorganic compounds, which serve myriad purposes. Even though these chemicals make up a small percentage of the liquid used for hydraulic fracturing, it can account for a median of over 10,000 kg (16) in the national average of 2.4 million gallons of water used per UOG well (17). This massive amount of chemicals is trucked to the pad site, stored and mixed onsite, and injected for hydraulic fracturing operations. Then, up to 30% of the water resurfaces during the flowback period before production begins. The storage, use, and collection of these chemicals, mixed hydraulic fracturing fluids, other chemicals involved in equipment cleaning and drilling processes, and the resulting flowback are all possible sources for groundwater contamination through controllable surface activities (18). Casing and cement failures (19) are a subsurface possibility for fluid introduction to the aquifer system, an event with little operator control, but which occurs at varying rates reported to be from 3% (20) up to as many as 12% of wells within the first five years (21).

	The majority of hydraulic fracturing additives can be found in lists that have been becoming more populated over the recent years. One of the earliest lists (22) was found in a report by the US House of Representatives Committee on Energy and Commerce. This included over 750 unique additives found in more than 2500 products available to be used for hydraulic fracturing from 2005 to 2009. Additional pertinent information included in the report are the number of products in which the compound is found, a table of additives that are health or environmental risks, and highlights of statistics for use of specific compounds of concern like 2-butoxyethanol. FracFocus (www.fracfocus.org), instated in April 2011, is the national hydraulic fracturing chemical registry. In the US, 28 states require chemical disclosure of hydraulic fracturing fluids, of which 22 are using FracFocus. It currently contains more than 80,000 disclosure documents from more than 1000 companies. This registry provides the operator, location, depths, chemicals, and mixed concentrations used in hydraulic fracturing activities (23). The companies that disclose this information are able to protect trade secrets with the ability to report some additives as “proprietary polymers” or under similar designations. A review of the FracFocus database from January 2011 through February 2013 (16) revealed more than 37,000 logs, which included chemical disclosure of 692 unique ingredients, of which 11% were deemed trade secrets.

	GC coupled to mass spectrometry (MS) has been the workhorse used to separate, detect, and possibly identify volatile and semivolatile compounds present in groundwater, after appropriate sample preparation (2). The MS detector is practically a requirement when surveying groundwater for contaminants related to UOG, even when performing targeted analysis for specific compounds. The qualitative information gained from the MS detector is invaluable in confirmation and unknown identification. The potentially complex mixture of compounds in groundwater impacted by UOG has generated false positives when using the suggested FID in EPA methods used for general groundwater (24).

	The overwhelming majority of the capability of unknown identification with GC–MS begins with the electron ionization (EI) source (1). The EI source generates diagnostic fragment ions of the compound in a systematic manner. The resulting spectra can then be matched across a number of mass spectral libraries, generated by the National Institute of Standards and Technology (NIST), National Institute of Health (NIH), and the EPA, among others. Another ionization source, chemical ionization (CI), can be used to complement the EI-resulting data. The molecular ion for the compound is generally preserved by CI. CI can be described as a softer ionization technique; therefore it generates fewer fragments and is not used as the primary source for unknown identification. An MS detector capable of CI can also possess the ability for negative chemical ionization (NCI), a selective ionization technique effective toward ionizing halogenated compounds. This selective ionization is a detriment to broad surveying of unknown compounds, but it is a valuable tool for researchers investigating halogenated species.

	Fragmentation information from tandem MS (MS-MS) for further confidence in identification can be generated by ion-trap or triple-quadrupole MS detectors. High resolution and accurate mass (HR-AM) analysis of ions for an additional identification vector can be achieved with time-of-flight (TOF) and orbital trap MS detectors. Hybrid MS detectors can also be found to combine the MS-MS capabilities with HR-AM on the back-end, as with a Q-TOF or Q-orbital trap.

	A portion of a Texas well water study conducted by our group included identification of the volatile and semivolatile compounds in groundwater across the Barnett Shale region, shown with sampling locations in Figure 2. A GC–MS method was developed to provide appropriate sensitivity and good sample throughput. The aim of the method was to extract and separate the greatest number of compounds from a small volume of water with minimal sample preparation. Sample preparation on the front end included a 2-mL ethyl acetate extraction from 5 mL of groundwater, shaken 1 min in a screw-top vial. GC–MS analysis was performed using a 30 m × 0.25 mm, 0.25-µm 

 Rxi-5ms (Restek) column with a single-quadrupole MS detector using an EI source. The “5” column, or 5% diphenyl, 95% PDMS stationary phase, is typically regarded as a general purpose column and has retention characteristics quite comparable to columns in which Kovat’s retention indices were calculated. The retention index of an unknown peak can be a valuable piece of data to narrow down possibilities of detected unknown compounds (25). Separation and MS parameters were set for the detection of 35 target compounds in our method. These compounds were chosen based on their popularity of use, possible health and environmental effects, detection in previous research, and GC amenability. These compounds consisted of various alcohols, aromatics, and other hydrocarbons. MS settings, summarized in Table I, included groups of selected ion monitoring (SIM) events for the base peaks of our target compounds, coupled with full spectral scanning for the confirmation of measured peaks by SIM, as well as the possible identification of unknowns through spectral matching. These acquisition groups were typically around 2 min each in an effort to keep the SIM ion count low to maintain an effective MS duty cycle. The scanning parameters also changed with each acquisition group, in that the concluding 

 increased from 100 to 400 over the time of the separation. The assumption used was that compounds eluted earlier from the column would be lighter than those eluted later; reducing the acquisition window reduces noise in the spectrum.  

	Initial application of this method detected methanol and ethanol in some groundwater samples. These detections were unable to be quantified with data at the time because of poor retention on the Rxi-5ms column and a fair amount of background noise from permanent gases like N

 This led our team to develop a method to address both of these problems.

	A mid-polarity GC column was chosen for the complementary analysis. The team still wanted to maintain the ability to adequately retain linear hydrocarbons if present, so working with a 100% PEG column was nearly out of the question, even though it maintains a great selectivity for these alcohols. The 30 m × 0.32 mm, 1.20-µm 

 Phenomenex ZB-BAC2 column, developed and marketed for blood alcohol analysis, was chosen for its retention and selectivity for these two alcohols and other solvents, along with the possibility of using a second paired column, the Phenomenex ZB-BAC1 column, for confirmation if needed. The method also incorporated FID to help reduce the background noise while detecting the light alcohols. A static headspace injection technique was chosen to effectively extract the analytes of interest and reduce background. A salt solution was added to the water sample to reduce the solubility of the alcohols in groundwater. Samples were agitated, heated, and injected automatically using an AOC-5000plus autosampler (Shimadzu Scientific Instruments).

	In our initial study of Barnett Shale groundwater in 2011 (26), 29 of the 100 samples contained methanol at concentrations as high as 329 mg/L and 12 samples contained ethanol at levels as high as 11 mg/L. These detections had no correlation with distance to UOG wells. Numerous industrial processes use these alcohols, and they can be produced through a range of biological pathways, so identifying the sources for the occurrences was not practical with the limited data.

	A follow-up study in 2014 expanded the research to 550 groundwater samples across 13 counties in north Texas (27), shown in Figure 2. Additional compounds were detected in this study in addition to the methanol and ethanol from the previous studies. Alcohols included methanol (35 wells), ethanol (240), isopropyl alcohol (8), and propargyl alcohol (155). Ethanol and propargyl alcohol had a positive correlation with each other. These are both ingredients in hydraulic fracturing fluids and were detected at a higher frequency than expected in the most productive counties based on chi-squared analysis. Chloroform, dichloromethane, and trichloroethylene were detected in 330, 122, and 14 wells, respectively. These chlorinated compounds are not disclosed ingredients in hydraulic fracturing fluid, but have been identified in UOGWW (28) and dichloromethane has been suggested (29) to be present during drilling operations as a degreaser for equipment. The study also found that 381 samples contained at least one aromatic of the BTEX class, with 10 samples containing all four species. Benzene was detected in 34 wells, toluene in 240, ethyl benzene in 22, and at least one xylene isomer in 240 water well samples. The BTEX compounds collectively can be found in hydrocarbon fuels, refined or unrefined, and some are used individually as industrial solvents, even as hydraulic fracturing additives.

	All of the compounds mentioned above can be linked directly or indirectly to UOG operations. The fact that these compounds are fairly common in the industrial or agricultural setting in which this research was conducted, renders it impossible to implicate UOG as the source of the contaminants with absolute confidence. It is expected that the only definitive manner in which to conclusively attribute UOG as a source of groundwater contamination would be through the detection of proprietary tracers (30), suggested to be fluorinated compounds exotic enough to assist each company with MS detection for internal monitoring.   

	Chromatography has been at the forefront of advanced analytical chemistry to tackle the challenges of analyzing complex mixtures related to UOG that possibly could be encountered during research. The previously discussed approaches are appropriate when identifying individual compounds, but there are situations when monitoring of bulk chemical classes yields adequate information. Many metals and ions can also be determined spectroscopically (31). For the most part, the operational costs for these methods are less than chromatography–MS methods, less technical to operate, and can even be performed portably. Spectroscopic approaches associated with UOG have included UV–vis spectroscopy, infrared (IR) spectroscopy, and optical emission spectroscopy (OES). Yet, many of these methods can fall victim to interferences from chemically similar compounds or ions since they are being measured in bulk solution without prior sample preparation. These methods are also typically intended for oil field waste waters or produced water, both of which commonly contain higher concentrations of the analyte than ever expected in compromised groundwater.

	Absorbance methods measured in the UV–vis region have been used for quantitating anionic surfactants (32), barium (33), boron (34), iron (35), sulfate (36), sulfide (37), and total petroleum hydrocarbons (TPH) (38). Anionic surfactants can be monitored near 655 nm as a complex with methylene blue according to EPA Method 425.1. Chloride solutions have been shown to give false positives with this approach (39). Barium has been measured as low as 2 mg/L as a precipitate after adding a sodium sulfate mixture and quantified at 450 nm. Strontium, silica, and calcium are the most detrimental interferences that may be encountered using this approach. Boron is measured at levels above 2 mg/L as the reaction product with carmine at 605 nm. Iron is monitored with the colorimetric phenanthroline indicator at 510 nm after the reagent has converted most forms of iron to a soluble ferrous iron. Iron can be measured down to 0.1 mg/L with the most common interference being a cumulative concentration of Ba

 greater than 50 mg/L. Sulfate at levels greater than 2 mg/L can be measured at 450 nm through a turbidimetric method after precipitation as barium sulfate. However, barium, magnesium, and silica present in the water sample can interfere with the accuracy of these results. Sulfide can be detected spectroscopically down to 0.01 mg/L at 665 nm after reacting with 

-phenylenediamine sulfate to form methylene blue. A semiquantitative method for TPH, which is a cumulative measurement of hydrocarbons ranging from C

, uses an immunoassay in which the hydrocarbons and enzyme compete to bind to antibodies immobilized on the cuvette. Measuring this absorbance at 450 nm yields a sensitivity equivalent to at least 2 mg/L diesel fuel. Chlorine present in solution can interfere with the assay.

	TPH can also be measured by IR spectroscopy. Previously, a method consisting of serial extractions with fluorocarbon-113 (CFC-113) and silica drying has been shown to be able to generate an extract adequate to quantify the absorbance of C-H stretches at 2950 cm

. Since CFC-113 is an ozone-depleting substance, the EPA has discontinued the method and suggests using the ASTM International Method D7006-04, which uses S-316 as a CFC substitute.

	OES has become the chosen technique for measuring dissolved metals in UOGWW (2). Metals of interest such as Ba, Sr, Fe, Na, Ca, and Mg are easily in the milligram-per-liter concentration range, comfortably above detection limits for inductively coupled plasma (ICP)-OES. The alkali and alkaline earth metals can be at levels of hundreds to thousands of milligrams per liter depending on the contribution of formation water to the overall UOGWW mixture. These excessive concentrations can be detrimental to an ICP-MS, accepted to be more sensitive than the ICP-OES. Most ICP-OES instruments come with the option to change between axial and radial viewing modes to assist in measuring samples across a wide concentration range, resulting in a much wider linear range for quantification than ICP-MS (40). Great attention needs to be taken in wavelength selection for the ICP-OES to ensure there is no spectral overlap from other metals at high concentrations or unexpected interferences from other hydraulic fracturing additives. Atomic absorption could also be implemented, but lacks the multielement throughput of ICP-OES.

A mix of analytical methods is required to understand the impact, if any, that UOG activity is having on groundwater.

Special Issues-10-02-2015

A Fast, Accurate, Speciation-Capable, Automated, and Green Gas-Phase Chemiluminescence Approach for Analyzing Waterborne Arsenic

GC–MS and UHPLC–MS-MS Analysis of Organic Contaminants and Hormones in Whale Earwax Using Selective Pressurized Liquid Extraction

The Benefits of Single-Particle ICP-MS to Better Understand the Fate and Behavior of Engineered Nanoparticles in Environmental Water Samples

Analytical Efforts Toward Monitoring Groundwater in Regions of Unconventional Oil and Gas Exploration

Fourier-Transform Molecular Rotational Resonance Spectroscopy: Bridging the Gap Between Spectroscopy and Chromatography for VOC Analysis

New Solutions to Environmental Challenges

The State of the Art of Flow-Through Solid-Phase Spectrometry