Infrared Spectroscopic Determination of Strongly Bound Cyanides in Water

Abstract

Cyanide species pose an environmental concern as they inhibit important biological processes in humans and aquatic systems. There is more focus on free-CN and weak acid dissociables cyanide as hazardous species compared to strong acid dissociables due to their higher reactivity and toxicity. However, the strong acid dissociables cyanide also poses health concerns as it liberates free-CN under ultraviolet irradiation or when present in acidic solutions. Detection of strongly acid dissociables cyanide typically requires its digestion in acidic solutions and measurement of the gaseous HCN produced. A simple infrared spectroscopic method is described here to speciate and quantify three strong acid dissociables cyanide: [Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻. The strategy involves precipitating the strongly acid dissociables cyanide using cetyltrimethylethylammonium bromide, capturing the precipitate on a polyethylene membrane, and quantifying the individual strongly acid dissociables cyanide from the IR spectrum recorded in transmission mode through the membrane. Controlling the particle diameter to be in the range of 0.2–2 µm is important. Particles less than 0.2 µm pass through the membrane, whereas particles larger than about 2 µm lead to nonlinearity in quantification. The average %recoveries for [Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻ were 100% (%RSD = 7), 91% (%RSD = 7), and 101% (%RSD = 8), respectively. The detection limit for [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻ were both 20 ppb CN⁻, whereas [Au(CN)₂]⁻ was 100 ppb CN⁻. The detection range was 20–750 ppb CN⁻ for [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻ and 100–750 ppb CN⁻ for [Au(CN)₂]⁻ with a linear regression of R² = 0.999–1.000.

1. Introduction

The total cyanide present in water is subdivided into three categories: (1) free cyanide (free-CN), (2) weak-acid dissociables (WADs-CN), which are cyanide ligands weakly bound to metal ions, such as Ag⁺, Cd²⁺, Cr²⁺, Cu²⁺, Hg²⁺, Mn²⁺, Ni²⁺, and Zn²⁺, and (3) strong-acid dissociables (SADs-CN), in which the cyanide ligands are strongly bound to metal ions, such as Au⁺, Co³⁺, and Fe³⁺ [1]. Free-CN toxicity to human health arises from its interaction with the ferric iron in metalloenzymes, inhibiting important biological processes [2]. The permissible limit for free-CN concentration in drinking water set by the Environmental Protection Agency (EPA) is 0.2 mg/L [3]. However, aquatic life has a lower tolerance toward free-CN, resulting in 22 µg/L as the EPA maximum level allowed in freshwater [4]. Free-CN exists predominantly as HCN (99.5%) in natural water at pH = 7, whereas the CN⁻ ion can form under alkaline conditions [5]. From a toxicity perspective, measurement of free-CN and WADs-CN is most important. However, SADs-CN do pose a hazard because the cyanide ions can be liberated from SADs-CN by photodissociation under sunlight [6]. In particular, the cyanidation process used to extract gold involves introducing large amounts of free-CN solutions, which can complex with other metals, creating SADs-CN. Therefore, they are particularly problematic in groundwater resources near gold mining areas. Currently, cyanide is being used in a “cyanide leaching” process in which gold is selectively leached out from ores forming [Au(CN)₂]⁻ (a SAD-CN), according to Equation (1). Gold is then recovered by adsorbing the [Au(CN₂)]⁻ solution onto activated carbon, which is then, in a second step, reduced back to Au⁰ [7].

2Au + 4NaCN + 2H₂O + O₂ → 2Na[Au(CN)₂] + 2NaOH + H₂O₂

(1)

In addition to forming [Au(CN)₂]⁻, cyanide-derived species found in gold mining include CNO⁻, CNCl, SCN⁻, Cu(CN)₃⁻, Zn(CN)₄²⁻, and Ni(CN)₄²⁻. Other SADs-CN, such as Fe(CN)₆³⁻ and Co(CN)₆³⁻ complexes, can also form as ore processing waste as a result of cyanide usage during the gold extraction process. Among SADs-CN, iron-based cyanides are the most common SADs-CN present in gold mining. Free-CN, however, does not persist in gold mining as it undergoes natural volatilization via HCN formation. Furthermore, the free-CN introduced to extract the gold oxidizes to CNO⁻ in the presence of certain oxidizers, which can be found in mining solutions such as oxygen, chlorine, ozone, and hydrogen peroxide. CNO⁻ does not tend to accumulate in underground water or gold mining regions since it hydrolyzes to ammonia and CO₂ [8]. Another factor that leads to conversion of free-CN to CNO⁻ is the presence of ferricyanide. Under alkaline conditions, ferricyanide undergoes reduction to ferrocyanide in the presence of free-CN, which is oxidized to the less toxic CNO⁻ [9,10]. Thus, the complex chemistry of cyanide makes it unlikely to find free-CN in gold mining regions [7].

Given the complex nature of the species of cyanide water, several EPA- and ASTM-approved methods are used to measure free-CN, WADs-CN, and the total CN (total-CN) concentrations. The most common of these methods are summarized in

Table 1. Standard methods of detecting free-CN, WADs-CN, and total-CN.

Type of Cyanide	Method	Description
Free-CN	ASTM 7237 [11]	Flow injection of a cyanide sample into a solution buffered with phosphate at pH range of 6–8; the generated HCN then diffuses through a gas diffusion membrane into an NaOH solution. CN− is released, and the anodic current is measured in an amperometric flowcell detector with a silver-working electrode.
Free-CN + WADs-CN	EPA 9016	Sample solution is buffered at pH 6 and introduced into a microdiffusion cell. The free cyanide diffuses as HCN, which then absorbs into a sodium hydroxide solution located at the center of the microdiffusion cell. The HCN solution is treated with acidified phosphate buffer and chloramine-T to convert the HCN to cyanogen chloride. The latter is reacted with pyridine-barbituric acid, forming a complex that absorbs at 578–587 nm.
Free-CN	ASTM 4282 [12]	Chlorination of free-CN with chloramine-T, followed by reaction with pyridine-barbituric acid; the resulting complex is measured colorimetrically, as described in EPA 9016.
WADs-CN	ASTM 6888 [13]	The sample is mixed with ligand exchange reagents followed by flow injection analysis. CN− is acidified to HCN and then diffuses through a gas diffusion membrane into an NaOH solution. The captured cyanide is measured amperometrically using a flow cell detector.
Total-CN	ASTM 7284 [14]	Sample is distilled in sulfamic acid containing magnesium chloride as the catalyst; CN− is acidified to HCN and measured amperometrically, as described in ASTM 6888.
Total-CN	ASTM 7511 [15]	Sample is digested with UV radiation to release CN−, followed by addition of concentrated sulfuric acid to form HCN. The resulting HCN is measured amperometrically, as described in ASTM 6888.
Total-CN	EPA 335.4 [16]	SADs-CN are converted to HCN gas by adding 18N sulfuric acid to the sample, followed by boiling and refluxing. The HCN is then collected in a scrubber containing sodium hydroxide solution. The HCN is then converted to cyanogen chloride by reacting it with chloramine-T, which subsequently reacts with pyridine and barbituric acid to give a red-colored complex. The complex is then measured colorimetrically at 570 nm.
Free-CN	APHA 4500-CN− method D [17]	Titration of free CN− against known amount of Ag+; the excess Ag+ reacts with p-dimethylaminobenzylidene rhodamine, forming a complex that absorbs at 490 nm.

.

Most of these methods involve acidification of the sample to convert all cyano complexes into HCN gas, which, in turn, is collected through a gas diffusion membrane into a sodium hydroxide solution to form free-CN. The amount of free-CN is then measured by means of colorimetry or amperometry. Sulfides are a major interference requiring the addition of lead to remove the sulfides prior to sample analysis [18]. The sample preparation procedure dictates the extent of conversion of the WADs-CN and SADs-CN into HCN and, hence, the level and type of cyanide detected. For example, total-CN is measured by EPA 335.4 and in ASTM standards D7284 and D7511. These methods involve digestion of the SADs-CN by UV or strong acids to release CN⁻ ions (see Table 1) and involve intensive labor work, use of strong acids, and generation of the highly toxic HCN gas.

An alternative approach for measuring free-CN that does not require formation and collection of HCN gas is APHA 4500-CN method D. In this case, the free-CN is titrated using silver nitrate to form Ag(CN)₂⁻ (a WAD-CN), and the concentration of the cyanide is determined by the endpoint at which excess Ag⁺ is detected. The excess Ag⁺ forms a complex with p-dimethylaminobenzylidene rhodamine, giving rise to a band at 490 nm (see Table 1) [17]. Since the free-CN is not directly measured but rather inferred by the fate of the Ag⁺ ion, this method is susceptible to matrix interferences. For example, the presence of anions like S²⁻ and S₂O₃²⁻ consumes some of the Ag⁺, forming Ag₂S and Ag₂(S₂O₃), which, in turn, produces a positive bias in the determination of the concentration of free-CN present in the sample.

Building on the Ag⁺ titration method, we recently developed an infrared spectroscopic-based method for measuring free-CN and WADs-CN that overcomes the issues associated with the Ag⁺ ion reacting with anions like S²⁻. In our approach, we do not titrate a sample with Ag⁺, but rather add an excess amount of Ag⁺ ions in one dose [19]. In this case, the excess Ag⁺ reacts with free-CN to form AgCN particulates, according to Equation (2). A known volume of the suspension is then passed through an IR-transparent membrane collecting the AgCN particles. The membrane is dried, and an IR spectrum is recorded directly through the Luer™ lock of the membrane assembly (Orono Spectral Solutions Inc, Hermon, USA). The amount of free-CN is determined from the intensity of the IR band due to the CN stretching mode at 2165 cm⁻¹. Sulfur-containing anions are not an interferent, as we directly measure a band due to CN⁻ rather than an indirect measure through the fate of the Ag⁺ ion. In our method, Ag⁺ does react with sulfur-containing anions such as S²⁻ to form Ag₂S particulates, and these too are collected on the membrane. However, the presence of AgS does not interfere with observing and measuring the intensity of the CN mode at 2165 cm⁻¹. In fact, adding S²⁻ to a sample is an advantage in our method as it enables measurement of the WADs-CN concentration. In essence, S²⁻ forms salts with the cations of the WADs-CN, releasing the CN⁻ ions, which, in turn, liberate free-CN (Equation (3)) that then reacts with Ag⁺ ions according to Equation (2).

AgNO₃ + CN⁻ AgCN + NO₃⁻

(2)

Cu(CN)₄²⁻ + S²⁻ CuS _(s) + 4 CN⁻

(3)

While developing a simple method to measure free-CN and WADs-CN was an important step forward, it was not a complete general method because it did not measure SADs-CN and, thus, total-CN. One could use EPA 335.4 or ASTM standards D7284 and D7511 to determine total-CN and, by difference, determine SADs-CN. However, this approach of combining multiple methods to measure free-CN, WADs-CN, and SADs-CN, in our opinion, is less attractive than establishing a general IR spectroscopic-based approach to measure all three types of cyanides and the total-CN.

To this end, we now describe a simple method to precipitate the three SADs-CN complexes of Fe(CN)₆³⁻, Co(CN)₆³⁻, and Au(CN₂)⁻ for collection on an IR-transparent membrane. Addition of CTAB results in precipitation of the Fe(CN)₆³⁻, Co(CN)₆³⁻, and [Au(CN₂)]⁻. While 100% recoveries were obtained for Fe(CN)₆³⁻ and Co(CN)₆³⁻, about 12% recovery was obtained for [Au(CN₂)]⁻ because of lower aggregation of the 1:1 Au(CN₂)⁻ CTAB complex. In this case, a 100% recovery of [Au(CN₂)]⁻ was achieved by spiking the solution with [Fe(CN)₆]³⁻. The precipitates of [Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN₂)]⁻ each produce unique CN bands, enabling both quantification and speciation of each SADs-CN. Furthermore, these bands do not interfere with the detection and measurement of the AgCN mode, enabling the determination of the concentration of free-CN, WADs-CN, and SADs-CN and, thus, establishing a general IR spectroscopic method for measuring total-CN.

2. Experimental

2.1. Materials and Reagents

Potassium hexacyanoferrate(III) (ACS reagent, 99.0%), potassium dicyanoaurate (I) (99.0%), potassium hexacyanocobaltate (98.0%), and CTAB were purchased from Sigma Aldrich (St. Louis, MI, USA). IR-transparent membranes (13 mm diameter, 0.1 µm pore size) were obtained from Orono Spectral Solutions Inc. (Hermon, ME, USA). Plastic Luer™ lock membrane assemblies were obtained from Millipore (Burlington, MA, USA) and modified, as described by Saleh and Tripp [20], to allow for the IR beam to pass directly through the Luer™ lock of the sample assembly. The assembly containing the membrane, a metal screen, and Teflon™ O-ring seals, as well as the procedure for passing a sample through the membrane, are described elsewhere [20].

2.2. Preparation of Cyanide Solutions

Stock solutions of 1000 ppm of K₃[Fe(CN)₆], K₃[Co(CN)₆], K[Au(CN)₂], and NaCN were prepared by dissolving 0.100 g of each cyanide compound in 100 mL of DI water. Then, a 2.5 mL aliquot of each of the 1000 ppm stock solutions was diluted with 50 mL with DI water to prepare 50 ppm stock solutions. These 50 ppm solutions were then used to make 50 mL working solutions with CN⁻ concentrations in the range of 2 ppm–50 ppb. Here, the concentration is given in terms of the CN concentration and not the concentration of the SADs-CN. To emphasize this distinction, [Fe(CN)₆]³⁻ solution containing 500 ppb CN⁻ is written as [Fe(CN)₆]³⁻ (500 ppb CN⁻).

2.3. Calibration of Cyanide Solutions Using ATR

A stock solution of 10,000 ppm was prepared from K₃[Fe(CN)₆] and K₃[Co(CN)₆] by dissolving 0.1 g of each material in 100 mL DI water. These solutions were used to generate 5000–10,000 ppm solutions for use in determining the extinction coefficient of each cyano species from Beer’s Law plots. For K[Au(CN)₂], a 30,000 ppm stock solution was prepared by dissolving 30 mg of K[Au(CN)₂] in 10 mL DI water. This stock solution was used to generate solutions in the range of 15,000–30,000 ppm. A drop of each solution was placed on a diamond crystal-ATR (PerkinElmer UATR two, Waltham, MA, USA), which fully covers the probed surface IR. Each spectrum was collected at 8 cm⁻¹ resolution and averaged from 10 scans. The ATR calibration plots are provided in the Supplementary Information.

2.4. Addition of CTAB

A 1000 ppm CTAB solution was prepared by dissolving 0.1 g of CTAB in 100 mL DI water. An appropriate volume was pipetted from the 1000 ppm CTAB stock solution and added in one shot into the 50 mL solutions containing [Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻ such that the total Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻:CTAB molar ratio was 1:10. The suspension was then stirred at 500 rpm for 30–45 min, followed by tip-sonication for 1 min at a frequency of 22.5 kHz using Fischer Scientific Sonic Dismembrator (model 100, Thermo Fisher Scientific, Waltham, MA, USA) to break up agglomerated particles. The pH measured for these complexes was in the range of 5.1–5.8, and addition of CTAB did not change the pH. We also prepared two control solutions in which one contained only [Fe(CN)₆]³⁻ (500 ppb CN⁻) and the other contained only CTAB (equivalent to the amount added in the [Fe(CN)₆]³⁻ (500 ppb CN⁻)) solution. DLS measurements were performed on a Malvern 3000 HAS Zetasizer (Malvern Panalytical Ltd., Westborough, MA, USA).

2.5. Environmental Samples

Lake water collected from Sabasticook Lake and tap water extracted from Orono’s municipal water in Maine were filtered through a 0.45 µm Teflon™ (Wilmington, DE, USA) to remove any suspended solid particulates [19]. Then, 1.750 mL of the 1000 ppm CTAB solution was added to 50 mL samples (9.6 × 10⁻⁵ M CTAB) of lake and tap water, and the same two matrices were spiked with either [Fe(CN)₆]³⁻-(60 ppb CN⁻) or [Co(CN)₆]³⁻-(60 ppb CN⁻). For measuring the concentration of [Au(CN)₂]⁻, 50 mL samples of lake and tap water were spiked with [Au(CN)₂]⁻-(300 ppb CN⁻) and [Fe(CN)₆]³⁻-(4.5 ppm CN⁻), followed by addition of 7.880 mL of the 1000 ppm CTAB solution (4.3 × 10⁻⁴ M CTAB).

2.6. Recording IR Spectra of Membranes Containing Cyanide Particles

The membrane was prewetted with a drop of ethanol prior to passing a sample suspension through the assembly. A syringe containing 10 mL of the suspension was attached to the Luer™ lock, and the suspension was then passed through the membrane vertically. The membrane was then gently air-dried by connecting the air tube through the neck of the Luer™ lock. Next, the Luer™ lock assembly was attached to an in-house IR sample holder, and an IR was spectrum-recorded through the Luer™ lock in transmission mode [20]. The spectra were recorded at 8 cm⁻¹ resolution with 2 times zero filling using an ABB-Bomem FTLA 2000 spectrometer (ABB-Bomem, Quebec City, QC, Canada) equipped with a DTGS detector. The assembly of the Luer™ lock containing the metal screen and the polyethylene membrane was used to record a reference spectrum. The CN⁻ band integrated intensity for each cyanide complex was determined using a valley-to-valley baseline correction.

3. Results and Discussion

3.1. Qualitative Aspects of the Method

3.1.1. Precipitation of SADs-CN Using CTAB

The use of CTAB to precipitate trivalent polyatomic anions for spectroscopic detection of ions was first reported by Asaoka et al. [21] to lower the detection limit of the molybdenum blue method for quantification of phosphate in water. CTAB was used to precipitate the phosphomolybdenum blue anion, [H₄PMo(VI)₈Mo(V)₄O₄₀]³⁻ PAMB). Then, the solid PAMB particles in suspension were captured on a membrane, followed by elution into a smaller volume of methanol, ethanol, or acetonitrile. This resulted in an increase in concentration by a factor of 10 of the PAMB in the solvent and, hence, a 10 times lowering of the detection limit [21]. Recently, we improved upon this approach by capturing the PAMB particles on a membrane transparent in the visible spectral region [22]. UV-Vis spectra were recorded directly on the membrane, avoiding the need for a solvent to elute the PAMB particles from the membrane back into solution. Thus, a natural extension would be to use CTAB to form particulates with the SADs-CN and capture them on an IR-transparent membrane for quantification by IR spectroscopy.

Figure 1a shows the IR spectrum after passing 10 mL of a suspension through an IR membrane that was generated by adding 0.584 mL CTAB to a 50 mL (3.2 × 10⁻⁵ M CTAB) solution containing 500 ppb CN⁻ of [Fe(CN)₆]³⁻ at a [Fe(CN)₆]³⁻:CTAB molar ratio of 1:10. The band at 2115 cm⁻¹ is due to CN⁻ stretching mode of [Fe(CN)₆]³⁻. Additional bands at 2920 and 2850 cm⁻¹ are the C-H asymmetric and symmetric stretching modes of CTAB, and the corresponding C-H bending modes are observed at 1470 cm⁻¹, showing that the particulates contained both CTAB and [Fe₃(CN)₆]³⁻. Artifacts appear in Figure 1b near 2920 and 2850 cm⁻¹ due to the strong underlying modes of the polyethylene membrane in this region. In a control experiment, no bands were observed when a 10 mL aliquot of a 50 mL solution containing 0.580 mL of the 1000 ppm CTAB stock solution (3.2 × 10⁻⁵ M CTAB) and 500 ppb CN⁻ of [Fe(CN)₆]³⁻ was passed through a membrane (Figure 1b,c). Furthermore, no particles were detected by DLS in these control samples, whereas particles of average particle diameter of 900 nm were obtained for the suspension in generating Figure 1a.

3.1.2. Speciation of Precipitated SADs-CN

A key pillar of our method is that a unique CN band is observed for each SAD-CN, which enables quantification of each individual SAD-CN. Figure 2a shows unique single CN stretching bands located at 2114, 2150, and 2128 cm⁻¹ in the IR spectra of the potassium salts of [Fe(CN)₆]³⁻, [Au(CN)₂]⁻, and [Co(CN)₆]³⁻ in KBr pellets, respectively. Three unique peaks are also observed in the IR spectra, shown in Figure 2b, for the precipitates collected on a membrane. A volume of 10 mL of each suspension was then passed through a membrane, and an IR spectrum was recorded. CN⁻ bands at 2140, 2123, and 2114 cm⁻¹ were observed for the precipitates of [Au(CN)₂]⁻, [Co(CN)₆]⁻, and [Fe(CN)₆]³⁻, respectively.

3.1.3. Particle Diameter

Controlling particle diameter is important for obtaining quantitative data. Particles that are less than 0.1 µm in diameter are too small to be collected on the 0.1 µm membranes. Our experience has shown that this occurs when the concentration of CN in the SADs-CN is low (CN < 100 ppb). In this case, we spike the sample with a known concentration of the SADs-CN to ensure the particle diameter is a minimum of 0.2 µm. On the other hand, particles that are too large (>2 µm) are equally problematic, as these particles can exceed the “attenuation length” or “absorption length” (L), creating opaque regions with a nonlinear response, unlike regions where the diameter < L, a linear response is observed (see Figure 3). Assuming linearity is maintained up to an absorbance value of 1, the adsorption length for each SAD-CN can be estimated using the extinction coefficient determined from ATR. As a first approximation of setting the particle diameter maximum to be equal the adsorption length and from the known density of each SAD-CN, we calculate the maximum particle diameter for [Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻, as 4.6 µm, 2.7 µm, and 1.8 µm, respectively. As a general practice, our experience shows that nonlinearity in detection with particle size begins to occur when the particle diameters exceed about 2 µm for the SADs-CN, which is consistent with calculated values for the maximum diameters. Therefore, the experimental procedures used to generate the precipitated SADs-CN by CTAB are selected such that the particles formed have diameters with values between 0.2 and 2 µm.

The problem with larger particle diameters is demonstrated in the Beer’s Law plot for precipitated particles of [Fe(CN)₆]³⁻ with CTAB generated using the [Fe(CN)₆]³⁻- (200 ppb–2 ppm CN⁻) (Figure 4a). In this case, a statistically weighted slope value of 72 ± 11 cm²/mg (R² = 0.946) was obtained. Figure 4b shows that increasing the concentration of suspensions of [Fe(CN)₆]³⁻ is associated with an increase in the particle diameter, particularly with particle diameter values reaching 19–25 µm at 1–2 ppm, which is above the calculated optical pathlength of ~2.3 µm for [Fe(CN)₆]³⁻. We then sonicated the suspension after addition of CTAB and found that the particle diameter dropped to the submicron level, making these values fall within the optical pathlength calculated above (Figure 4b). The IR spectra recorded for these sonicated suspensions, after being passed through a membrane (Figure 4c), showed an increase in slope value 123 ± 3 cm²/mg (R² = 0.998). Thus, as a general requirement, the suspensions are sonicated before extracting a sample into the syringe.

3.2. Quantitative Aspects of the Method

3.2.1. Quantification of [Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻ on a Membrane

Figure 5a,b is plots of the concentration of CN⁻ from [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻ determined by the IR method versus the known concentration in solution. The ordinate values were generated by passing 10 mL of the suspension obtained from adding 0.120–1.750 mL CTAB (6.6 × 10⁻⁶–9.6 × 10⁻⁵ M CTAB) to 50 mL solutions containing 20 ppb–750 ppb of [Fe(CN)₆]³⁻ or [Co(CN)₆]³⁻. Details for calculating the concentration of the target analyte are described elsewhere [19,20] and are provided in the Supporting Information (Section S1.1., Figure S1). We note that the calculated concentrations from the IR spectra are not derived from a calibration plot but rather from first principles using known extinction coefficients. Thus, the slopes of the curves in Figure 5a,b are true %recoveries. The slopes in Figure 5a,b correspond to 100% (%RSD = 7%) and 91% (%RSD = 7%) recoveries for [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻, respectively. The limit of detection (LOD) was calculated using EPA guidelines by multiplying the standard deviation for the lowest concentration in our Beer’s Law plots × 10. This gives an LOD of 20 ppb CN⁻ for both [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻ and 100 ppb CN for [Au(CN)₂]⁻. However, a similar plot to those shown in Figure 5 generated for solutions containing [Au(CN)₂]⁻ gave a %recovery of only 12%, which we attribute to a lower level of precipitation [Au(CN)₂]⁻ with the CTAB.

The low level of [Au(CN)₂]⁻ precipitation with the CTAB can be explained on the basis of the number of CTAB molecules surrounding [Fe(CN)₆]³⁻ or [Co(CN)₆]³⁻ versus [Au(CN)₂]⁻. An estimate of the CTAB:[Fe(CN)₆]³⁻ mole ratio in the precipitate is obtained from the relative intensity of the band at 2850 and 2112 cm⁻¹, corresponding to v₁(C-H) in CTAB and v₂(C-N) in CN⁻. The 2850 and 2112 cm⁻¹ band ratios were first calibrated using a physical mixture containing known quantities of K₃[Fe(CN)₆] and CTAB pressed in NaCl pellets. Using this calibration, we determine that there are 3 molecules of CTAB surrounding each molecule of [Fe(CN₆)]³⁻. Similarly, the CTAB:[Co(CN)₆]³⁻ band ratio indicates the same number of CTAB molecules surrounding the trivalent [Co(CN)₆]³⁻; i.e., 3–4 CTAB molecules per complex. It is the combination of charge neutralization and hydrophobic effect of the alkyl tails of the CTAB that leads to aggregation and precipitation of the SADs-CN anions (see pathway (a) in Figure 6). On the other hand, [Au(CN)₂]⁻ is neutralized by 1 CTAB molecule. This lower number of CTAB molecules leads to a low level of aggregation needed for particle growth. To improve the %recoveries, the [Au(CN)₂]⁻ is co-precipitated with [Fe(CN)₆]³⁻ or [Co(CN)₆]³⁻ along with CTAB. Each CTAB molecule neutralized by [Au(CN)₂]⁻ aggregates with the more hydrophobic [Fe(CN)₆]³⁻ or [Co(CN)₆]³⁻ surrounded by 3 CTAB molecules (see pathway (b) in Figure 6). When spiking solution containing [Au(CN)₂]⁻ with [Fe(CN)₆]³⁻ before addition of CTAB results in co-precipitation of [Au(CN)₂]⁻ and %recoveries of 100%. Recall that the precipitates contained approximately 3 CTAB molecules per each of the trivalently charged [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻ ions. Charge neutralization would occur with both [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻, and this, along with the increased hydrophobicity of the alkyl tails, leads to aggregation and precipitation of these species. However, the ratio between [Au(CN)₂]⁻ and CTAB is only 1:1, leading to a lower level of aggregation and precipitate formation. To achieve a higher level of precipitation and, hence, a higher %recovery of the [Au(CN)₂]⁻, we spiked the solutions with [Fe(CN)₆]³⁻ to create nucleation sites that enabled the precipitation of all [Au(CN)₂]⁻. In this case, [Au(CN)₂]⁻ neutralized by 1 CTAB molecule undergoes tail-to-tail hydrophobic interactions with the CTAB molecules surrounding [Fe(CN)₆]³⁻ and, hence, co-precipitate [Au(CN)₂]⁻.

Table 2. %Recovery of [Au(CN)₂]⁻-(500 ppb CN⁻) when spiked with 2.0, 2.5, and 3.0 ppm CN⁻ of [Fe(CN)₆]³⁻.

[Fe(CN)6]3− Spiked	%Recovery of [Au(CN)2]−	% (±SD) (n = 3)	%Recovery of [Fe(CN)6]3−	% (±SD) (n = 3)
0	12%		-
2.0	84%	1.5	98%	1.7
2.5	82%	8.5	97%	10.5
3.0	95%	6.8	93%	2.7

shows the results using three solutions each containing [Au(CN)₂]⁻-(500 ppb CN⁻), spiked such that the final concentrations of [Fe(CN)₆]³⁻ in the 50 mL sample are 2.0, 2.5, and 3.0 ppm CN⁻. The amount of [Au(CN)₂]⁻ recovered increased to 95% when spiked with [Fe(CN)₆]³⁻-(3.0 ppm CN⁻) (Table 2), which was then used to generate the plot shown in Figure 7. On the other hand, the %recoveries for [Fe(CN)₆]³ were in the range of 93–98% for all concentrations of CN⁻ between 2.0–3.0 ppm with SDs in the range of 1.7–10.5%. We previously showed in Figure 6 that the −3 charge on [Fe(CN)₆]³⁻ results in neutralization by 3 CTAB molecules and results in aggregation and precipitation of [Fe(CN)₆]³⁻ without being co-precipitated with another SAD-CN. The data represented in Table 2 corresponds to experiments done in triplicate (n = 3) in which 10 mL aliquots withdrawn from the same beaker were passed through three different membranes.

Figure 7 is a plot of the CN⁻ concentration determined by our IR method using [Au(CN)₂]⁻–(100–750 ppb CN⁻) precipitated with [Fe(CN)₆]³⁻-(1.5–4.5 ppm CN⁻) and (2.6–7.880 mL CTAB, equivalent to 1.4 × 10⁻⁴–4.3 × 10⁻⁴ M CTAB) (See Section S1.2. in Supporting Information for further details) versus the known concentration of CN⁻ in solution. The slope shows a 101% (%RSD = 8%) recovery for [Au(CN)₂]⁻.

3.2.2. Mixture Containing [Au(CN)₂]⁻, [Fe(CN)₆]³⁻, and [Co(CN)₆]³⁻

The next step was to test the method with mixtures of all three SADs-CN. Mixtures containing 500 ppb CN⁻ from each of [Au(CN)₂]⁻, [Fe(CN)₆]³⁻, and [Co(CN)₆]³⁻ were precipitated with CTAB at [complex]:[CTAB] molar ratio of 1:10. This enabled the measurement of the concentrations of [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻. To measure [Au(CN)₂]⁻, a second sample solution was also spiked with [Fe(CN)₆]³⁻-(3 ppm CN⁻). The IR spectra obtained for these two cases are shown in Figure 8. In Figure 8a, the CN⁻ bands for [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻ at 2115 and 2126 cm⁻¹ overlap. We found that by spiking with 3 ppm CN⁻, the intensity of the lower frequency band at 2115 cm⁻¹ increases without overlapping with [Au(CN)₂]⁻ band at 2140 cm⁻¹ (Figure 8b). The small band appearing at 2100 cm⁻¹ is a second asymmetric CN stretching mode of [Fe(CN)₆]³⁻. Thus, to determine the intensities of each band in Figure 8b, we performed a scaled subtraction using the spectrum obtained for [Co(CN)₆]³⁻-(500 ppb CN⁻) to remove the contribution of the 2126 cm⁻¹ from the spectrum. By multiplying the scalar value used in the subtraction procedure by 500 ppb and adjusting for any volume difference passed through the membrane, the concentration of the [Co(CN)₆]³⁻ was determined. For the [Fe(CN)₆]³⁻, the amount was determined from the integrated intensity of the band at 2115 cm⁻¹ in the subtracted spectrum. Similarly, the concentration of [Au(CN)₂]⁻ was determined using the integrated area under the peak at 2140 cm⁻¹ in the spectrum in Figure 8b after subtraction. The band due to [Fe(CN)₆]³⁻ did not overlap with the CN band of [Au(CN)₂]⁻ at 2140 cm⁻¹, and thus, no scaled subtractive procedure was needed to determine the integrated intensity of this band. The resulting %recoveries and standard deviations (n = 3) obtained were 102% ± 5%, 98% ± 5%, and 100% ± 4% for [Co(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻, respectively.

3.3. Environmental Samples of Unknown Concentrations

The flow chart of the approach used for running samples with unknown concentrations of SADs-CN is shown in Scheme 1. The first step is to add 1.750 mL from 1000 ppm CTAB stock solution to a 50 mL sample (9.6 × 10⁻⁵ M CTAB) and process 10 mL of the suspension through a membrane. This amount of CTAB was chosen because it ensures a factor of 10 higher concentration than the SADs-CN at the high limit of the detection range. If bands at 2115 or 2126 cm⁻¹ are observed, then the concentration of [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻ is determined and in the range of 100 to 750 ppb. If no bands at 2115 or 2126 cm⁻¹ are detected, then 30 mL of the suspension is passed through a second membrane. If bands at 2115 or 2126 cm⁻¹ were detected, then the concentrations of [Fe(CN)₆]³⁻ and [Co(CN)₆]³⁻ are in the range of 20 to 100 ppb. Alternatively, one sample could be processed using a 30 mL sample to cover the entire detection range. The same flowchart is used for measuring the level of [Au(CN)₂]⁻⁻ in a sample with one additional step. In this case, 0.530 mL from 1000 ppm CN⁻ stock solution of [Fe(CN)₆]³⁻ is added to a 50 mL sample (to make a final concentration of [Fe(CN)₆]³⁻-(4.5 ppm CN⁻) in the sample) before proceeding through the flowchart.

Common cations present in environmental water samples are Na⁺, Ca²⁺, Mg²⁺, and K⁺, whereas common anions are Cl⁻, HCO₃⁻, CO₃²⁻, and SO₄²⁻ [23]. Thus, environmental samples obtained from freshwater presented herein test our method in the presence of these ions. Furthermore, precipitates of other common ions with CTAB may occur, but if formed, would not have IR bands that are not located in the region containing the CN bands. Samples obtained from tap water and Sabasticook Lake were analyzed according to the flowchart in Scheme 1 for the presence of [Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻. A non-detect of all three SADs-CN species was obtained, indicating that the SADs-CN, if present, were in concentrations below the detection limit or that an interferent was present in the lake and tap waters that prevented particulate formation. Recoveries with matrix spikes were then used to determine if an interferent was present in the lake and tap waters.

Water samples were spiked with [Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻ at the 60 ppb and 200 and 300 ppb levels. %Recoveries ranged from 81% to 108%, with an average 95% recovery and 5% RSD over all samples tested (

Table 3. CN⁻ concentration, %recovery, and %RSD for matrix spikes of [Fe(CN)₆]³, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻ into the Sabasticook Lake and Orono Municipal water.

Cyanide Species Spiked	Recovered CN− (ppb)	%Recovery	% (±RSD) (n = 3)
Sabasticook Lake
[Fe(CN)6]3−-(60 ppb CN−)	63	105	4
[Co(CN)6]3−-(60 ppb CN−)	65	108	4
[Au(CN)2]−-(300 ppb CN−)	244	81	5
Tap water
[Fe(CN)6]3−-(60 ppb CN−)	51	86	5
[Fe(CN)6]3−-(200 ppb CN−)	189	94	5
[Co(CN)6]3−-(60 ppb CN−)	58	96	10
[Au(CN)2]−-(300 ppb CN−)	290	97	5

). This shows that there was no matrix interference in these representative environmental samples. It is also noted that the IR spectrum provided insight into the lowest %recovery value of 81% for 60 ppb [Fe(CN)₆]³⁻ in tap water. The lower %recovery for [Fe(CN)₆]³⁻ is attributed to a reduction of ferricyanide to ferrocyanide in this matrix by evidence of the band appearing at 2070 cm⁻¹ [24,25]. This shows the additional advantage of an IR method to speciate between different types of cyanides and investigate the complex chemistry of cyanide in water.

4. Conclusions

A method for measuring the concentration of SADs-CN by IR spectroscopy has been demonstrated. Current methods require digestion of the SADs-CN with highly concentrated sulfuric acid and capture of the generated HCN gas. Our method avoids both the use of concentrated acids and evolution of a highly toxic gas. In our method, the SADs-CN form precipitates with CTAB that are captured on an IR-transparent membrane, followed by recording of an IR spectrum in transmission mode through the membrane. Sonication of the suspensions is important to maintain the particle diameter below the absorption length to achieve linearity across the detection range of 20 ppb to 750 ppb. The %recovery for [Fe(CN)₆]³⁻, [Co(CN)₆]³⁻, and [Au(CN)₂]⁻ were 95 ± 7%, 92 ± 7%, and 100%, respectively, with a high linearity (R² = 0.995, 0.999, and 1.00, respectively). Matrix spikes in lake and tap water showed an average 95% recovery and 5% RSD over all samples tested.