Assessment and Correction of Turbidity Effects on Raman Observations of Chemicals in Aqueous Solutions Joseph V. Sinfield,a,* Chike K. Monwubab a b

School of Civil Engineering, Purdue University, West Lafayette, IN 47907 USA Environmental Resource Engineering Department, Humboldt State University, Arcata, CA 95521 USA

Improvements in diode laser, fiber optic, and data acquisition technologies are enabling increased use of Raman spectroscopic techniques for both in lab and in situ water analysis. Aqueous media encountered in the natural environment often contain suspended solids that can interfere with spectroscopic measurements, yet removal of these solids, for example, via filtration, can have even greater adverse effects on the extent to which subsequent measurements are representative of actual field conditions. In this context, this study focuses on evaluation of turbidity effects on Raman spectroscopic measurements of two common environmental pollutants in aqueous solution: ammonium nitrate and trichloroethylene. The former is typically encountered in the runoff from agricultural operations and is a strong scatterer that has no significant influence on the Raman spectrum of water. The latter is a commonly encountered pollutant at contaminated sites associated with degreasing and cleaning operations and is a weak scatterer that has a significant influence on the Raman spectrum of water. Raman observations of each compound in aqueous solutions of varying turbidity created by doping samples with silica flour with grain sizes ranging from 1.6 to 5.0 lm were employed to develop relationships between observed Raman signal strength and turbidity level. Shared characteristics of these relationships were then employed to define generalized correction methods for the effect of turbidity on Raman observations of compounds in aqueous solution. Index Headings: Raman spectroscopy; Time-resolved Raman spectroscopy; TRRS; Turbidity correction; In situ.

INTRODUCTION Quantitative optical spectroscopic analysis of aqueous solutions in or extracted from the natural environment has historically suffered from interference associated with a variety of factors including fluorescence from natural and anthropogenic fluorophores, biofouling, complex non-target chemical backgrounds, and the presence of microorganisms.1–3 Significant research has been carried out to explore and develop management methods for many of these issues (e.g., management, suppression, and avoidance of fluorescence). While several of these challenges remain less than perfectly addressed (e.g., biofouling), even early work has helped to make application of spectroscopy in the field more commonplace and practical.2,4–9 The intent of the work presented herein is to investigate yet another key source of interference Received 16 September 2013; accepted 8 July 2014. * Author to whom correspondence should be sent. E-mail: jvs@purdue. edu. DOI: 10.1366/13-07292

Volume 68, Number 12, 2014

encountered in the field—turbidity. Turbidity is a measure of the loss of optical transparency of a medium resulting from the presence of suspended solids or other interfering matter, which can limit the overall sensitivity of optical spectroscopic methods and make it challenging to perform quantitative analysis. This adverse influence on optical spectroscopic measurements is recognized in many fields.

PERSPECTIVES ON TURBIDTY In oceanographic investigations, the challenges posed by turbidity for in situ spectroscopic analyses have been acknowledged for several decades. Ivanon et al.10 indicated that turbidity limited the effectiveness of studies of petroleum films, chlorophyll distribution, and sea water temperature and salinity performed using airborne laser sounding. In a similar domain, Phillips11 and Hoge and Swift12 noted a reduction in the Raman backscatter of water with increasing turbidity in airborne examinations of water optical transmission. Further, Amann13 called attention to the fact that turbidity can adversely affect the potential to measure low concentrations of natural and hazardous chemicals in the ocean. Other researchers14,15 more deeply explored the mechanisms underlying turbidity’s impact on spectroscopic observations of systems involving natural waters and tracers and distinguished the effects of scattering and absorption. In addition, in the pharmaceutical, biological, and medical fields, research has indicated that sample turbidity—in this case resulting from inhomogeneities in tissues and fluids—adversely influences Raman analyses of bone, blood, and tissue. Vellekoop and Aegerter16 indicated that in vivo studies involving fluorescent markers are limited by turbid tissue constituents that scatter light and prevent it from reaching the image plane of fluorescence microscopes. Further, Schulmerich et al.17 and Barman et al.18 noted the challenges of turbidity encountered in transcutaneous Raman studies of bone and tissue, respectively. Other researchers19–21 have examined the unfavorable influence of turbidity on Raman observations in stratified turbid media such as pharmaceutical tablets and pharmaceuticals contained within plastic packaging and diffusely scattering plastic bottles. In the context of environmental science, several researchers have indicated that the presence of suspended particulates in open fresh water and groundwater can lead to scattering, absorption, and displacement of an unknown sample volume that would otherwise be

0003-7028/14/6812-1381/0 Q 2014 Society for Applied Spectroscopy

APPLIED SPECTROSCOPY

1381

interrogated by optical sources in a spectroscopic system and can thus lead to inaccurate measurements of in situ chemical concentrations. Braunlich et al.22 initially noted the adverse impact of turbidity in some of the first laser-based Raman sensing studies of pollution in natural water. Over time, other researchers2,4,23,24 explored the significance of the displaced volume and absorption effects of particulates in greater detail highlighting reductions in both Raman and fluorescence system sensitivity for optical spectroscopic measurements performed on fluids in granular media relative to those performed on the same liquid analyte concentrations in solutions containing no particles. In 2004, Langergraber et al.25 called out light scattering and absorption effects in ultraviolet–visible (UV-Vis) spectroscopic studies of paper mill wastewater. Most recently, Downing et al.26 quantified the effect of turbidity on fluorescent dissolved organic matter observations in water, showing almost 90% signal attenuation at a turbidity level of 600 formazin nephelometric units (FNU).

HISTORICAL ATTEMPTS TO CORRECT FOR TURBIDITY’S INFLUENCE Given the inherent value of performing spectroscopic measurements in situ, and the limiting effects of turbidity, a great deal of effort has also gone into attempts to correct for its impact on spectroscopic observations. The approaches pursued depend significantly on the purpose of the performed measurement. Some researchers desire to see ‘‘through’’ the turbid medium to perform spectroscopic analysis on underlying layers of material. This is the case in many biological scenarios and studies of pharmaceuticals. In these contexts, researchers have developed techniques to either (a) work with the limited amount of unaffected light returned from a turbid sample (e.g., confocal microscopy or multiphoton microscopy)27 often at considerable complexity and expense, (b) invert the scattered return, sometimes employing sophisticated time gating, with tradeoffs in resolution,28–30 (c) peer ‘‘through’’ turbid layers via what has been termed interferometric focusing,16,31,32 with the limitation of a need for a priori access to the target focal plane, and/or (d) employ non-linear chemometric or bioinformatic methods to address nonanalyte-specific signal variances stemming from turbidity through techniques such as support vector regression, after considerable experimental trials or work with simulated samples.33 Others have focused on studying the bulk turbid material itself—that is, the matrix responsible for generating the turbidity effect in the sample—which in many applications is granular or powdery in nature (e.g., bulk active pharmaceutical ingredient analysis). Here again time-resolved scattering observations can be employed to discern the source of scattered return as a function of the probable scattered photon path length,19,29,34 or a spatial offset between the incident illumination source and the collected return can be employed19,21 to reassemble an image from diffusely scattered light. In addition, in related areas of work, researchers have highlighted the merits of employing

1382

Volume 68, Number 12, 2014

a large interrogation spot size and making use of reflecting mirrors or diffuse reflectors to enhance Raman returns when the scattering medium is of analytical interest.35 Still other researchers perform analyses of liquid (water) or tissue samples and wish to study the chemical composition of the fluid or tissue via quantitative analysis of the medium in which the turbid inducing constituents are suspended (more narrowly, the Raman scatterers within the turbid medium). In these contexts, differing methods have been employed for different target materials, with notable variations in approach between the laboratory and the field. For analysis of biological tissues in the laboratory, where turbidity correction is arguably the most advanced, corrections were initiated in the context of fluorescence studies. Here researchers recognized that fluorescent and diffusely reflected photons behave in a similar manner in turbid media, and methods were derived to obtain what is termed ‘‘intrinsic fluorescence’’, that is, the fluorescence emanating from direct excitation incidence on the target from either the ratio of measured fluorescence to diffuse reflectance at a given emission wavelength34,36,37 or through interpretation of photon migration models of concomitantly measured fluorescence and reflectance.38 A similar line of logic was employed by researchers39,40 in the development of intrinsic Raman spectroscopy (IRS) and turbidity-corrected Raman spectroscopy (TCRS),18 which are both based on the photon migration approach and employ alternate acquisition of Raman and diffuse reflectance spectra to obtain corrected Raman observations. The IRS method employs a Monte Carlo calibration model based on extensive analysis of phantom media representative of target constituents, as well as accurate knowledge of constituent Raman scattering coefficients. The TCRS method18 overcomes these limitations through a theoretical link between the observed Raman spectrum, the diffuse reflectance spectrum, and the turbidity-corrected Raman spectrum. The method, however, requires complex calibration to obtain an instrument specific constant, determination of the average photon path length in the turbid media under investigation, and estimation of the diffusely reflected light at the Raman excitation frequency. Reble et al.41 also suggest turbidity corrections based on the relationship between Raman return and sample reflectance, but distinguish the contributions of absorption and scattering and employ a Monte Carlo simulation to obtain a corrected Raman signal from an inferred combination of the target material absorption coefficient and reduced scattering coefficient. Although complex and challenging to implement in a natural field setting, the breakdown of the influences of turbidity on Raman observations provided by these model-based corrections reveals that for quantitative characterization of turbid media, turbidity-induced variations in sampling volume—that is, the turbidity-inducing constituents occupy a fraction of the interrogated sample volume—often become dominant over other forms of spectral distortion (absorption and scattering).39,42

This premise has historically been exploited in the natural environment where turbidity corrections have been primarily linked to direct or indirect measurements of sample turbidity that provide an indication of combined scattering, absorption, and volume reduction effects. Tipton and Vogt43 corrected measurements of fluorescent tracers in model aquifers for turbidityrelated signal attenuation through cyclic excitation of target media and a known fluorescence reference positioned on the far side of a flow through sampling vessel, enabling real time correction for signal amplitude, with the known drawback of potential interference from fluorophores within the target media. Similarly, Schnegg15 put forward a simple groundwater fluorometer signal intensity correction based on an empirically derived relationship between sample turbidity level and related changes in scattered excitation energy. Langergraber et al.25 corrected UV-Vis absorption measurements of paper mill wastewater for turbidity by a theoretical relation linking turbidity to light attenuation based on an assumption of the diameter of the turbidityinducing particles and a spectral correction relating changes in scattering at any given wavelength to turbid particle diameter. More recently, Downing et al.26 demonstrated the potential to correct observations of fluorescing dissolved organic matter for turbidity effects through instrument-specific calibration of excitation light absorption as a function of dissolved organic matter content in target media, and empirically derived relations between field-derived turbidity levels and fluorescence signal attenuation, noting the need for periodic collection of water samples and measurement of filtered samples in the laboratory to assess the appropriate magnitude of the correction. Despite these more recent advances, perhaps one of the most intriguing corrections for turbidity stems from work by Bristow et al.,44 who utilized the water O–H stretch Raman emission to correct airborne laser fluorosensor observations of phytoplanktonic chlorophyll a for effects of turbidity-induced water optical attenuation. The reciprocal of the water Raman signal was found to vary directly with the beam attenuation coefficient. No corrections are known to be published for Raman observations of turbid aqueous samples in the natural environment.

TURBIDITY CORRECTION IN ENVIRONMENTAL FIELD MONITORING With this background in mind, this paper presents a study of the influence of turbidity level and turbiditygenerating particle size on quantitative Raman measurements of water contaminated with common environmental pollutants and puts forward potential means to correct for these effects that can be applied in a typical field monitoring scenario. This application places important constraints on any approach to turbidity correction in that environmental applications of Raman spectroscopy are often performed down-hole in a monitoring well, through an optical window in direct contact with soil beneath the ground surface via an in situ probe, or in a submersible unit deployed in open water. The measurements therefore typically

employ a 1808 backscatter, single-sided acquisition geometry (limiting the potential for ‘‘pass-through’’ observations). In addition, the conditions monitored, especially when probing the ground or in open water, are constantly changing, making use of a priori modeling impractical. Further, the depth of penetration of light into what is often very turbid media, or water within the pore space of soil, is often quite limited. Beyond this, the desire to observe chemicals in the water at relatively low concentrations (on the order of parts per million to parts per billion) and ultimately to limit interference from naturally occurring fluorophores encourages use of a small irradiated spot size to both (a) maximize the energy incident on the target per unit area (with the area often constrained by the pore size of a soil matrix) and (b) limit the return path length of the detected energy to reduce convolution of Raman and fluorescence returns and enable timeresolved analysis. Turbidity correction is therefore primarily driven by absorption and volume reduction effects and is likely most effective when performed in real-time, based on spatially relevant in situ turbidity information. Herein, two in situ indicators of the optical properties of the studied solutions are therefore explored as the basis for potential turbidity corrections, each of which is examined in the context of a specific contaminant commonly encountered in the environment. First, the Raman return associated with the OH stretching band of water is examined as a direct indicator of the effect of turbidity on Raman signatures. Changes in this return are employed to define a turbidity correction for solutions containing strong scatterers that have little influence on the Raman spectrum of water, in this case ammonium nitrate, which is typically encountered in the runoff from agricultural operations. Next, a direct measure of turbidity is employed as the basis of a correction for solutions containing compounds that have a notable impact on the Raman spectrum of water, which herein are represented by trichloroethylene, a pollutant commonly encountered at contaminated sites associated with degreasing and cleaning operations. Both cases are described in detail below, with associated turbidity corrections developed through a parametric investigation of turbidity level and particle size performed using aqueous solutions doped with silica flour particles that remain suspended in the aqueous test medium for the duration of a spectroscopic observation.

MATERIALS AND METHODS The materials employed in this study included both chemicals and soil grains. Chemicals. As noted above, two compounds were examined as target analytes in laboratory-prepared aqueous solutions: ammonium nitrate and trichloroethylene (TCE). Ammonium nitrate, which has little impact on the Raman vibrational modes of water, was used as a relatively innocuous contaminant analog in extensive parametric studies of the effect turbidity level and turbidity inducing soil grain size on analyte Raman observations. The ammonium nitrate employed

APPLIED SPECTROSCOPY

1383

in this study was obtained from VWR (Product no. BDH0212, ACS Grade, 95%). TCE, a chlorinated hydrocarbon and member of a group of compounds known as dense nonaqueous phase liquids, was employed in a narrower set of tests as an analyte known to influence the Raman vibrational modes of water.45,46 A spectrometric grade (purity . 99.5%) of the compound obtained from Alfa Aesar was used as received for this work. Soil Grains. In order to minimize the influence of chemical or physical interactions that might occur between the turbid-inducing media employed in this study and the compounds in aqueous solution, relatively inert homogenous silica flour was used to create the turbid samples (CAS No. 14 808-60-7). Multiple Raman analyses of aqueous solutions of nitrate and trichloroethylene containing either 1.6 or 5.0 lm silica flour at varying turbidity levels were performed at discrete time intervals over extended periods of 0–24 h to evaluate any potential adsorption of the target analytes by the silica particles. One set of tests was performed with shaking before each measurement to ensure that the introduced particles were in suspension during testing, while a second set of tests was performed without shaking after introduced particles were allowed to settle. Results of the tests displayed no variations in target line Raman intensities (the NO3 symmetric stretching peak at 1067 cm1 for ammonium nitrate, and the 381 cm1 (dCH) Raman shift of TCE) over time beyond those observed during repeated tests performed immediately after solution preparation, indicating that chemical adsorption on the silica, if present at all, was not significant. These findings are consistent with what would be anticipated based on literature on the adsorption characteristics of silica.47 The material is white to tan in color, has a specific gravity of 2.65, and is approximately 99.0–99.9% SiO2, ,0.8% Al2O3, ,0.1% Fe2O3, ,0.1% TiO2 by weight. This material, which was commercially sourced from AGSCO Corporation (Wheeling, IL), had been sieved to obtain distinct samples, each with a well-defined grain size distribution containing at least 80% of one of the following grain sizes: 1.6, 2.1, 2.4, 3.5, and 5.0 lm, as specified by the manufacturer. When employed as suspended particles in aqueous solutions in the tests described below, these samples are referred to simply by their prominent grain size. Aqueous Solution Preparation. Ammonium Nitrate. Aqueous samples of nitrate solutions were prepared by adding the desired amount of reagent grade ammonium nitrate (NH4NO3) by weight to a single liter of de-ionized water and mixing thoroughly. All solution concentrations were verified by copper–cadmium reduction. Trichloroethylene. Aqueous solutions were prepared in a 40 mL amber-colored glass vial topped with a Teflon-coated silicone rubber septa sealed and screwed in place with a cap. All preparations were undertaken at room temperature (24.5 8C). The 40 mL bottles were filled with deionized water to a capacity of 20 mL, and then neat TCE was placed at the base of the vials through the use of a disposable pipette. This helped to ensure that no bubbles were introduced in the process, which was stopped when the supernatant water was overflowing. The vials were then capped and sealed immediately,

1384

Volume 68, Number 12, 2014

avoiding headspace and preventing loss due to vaporization. These vials were then shaken and allowed to settle to equilibrium for at least three days. To prepare dilute concentrations, aliquots were taken through the use of a syringe and injected into a closed vial containing deionized water in an appropriate volume needed to make a particular concentration based on a mass balance calculation. Final concentrations were validated using a gas chromatography dry electrolytic conductivity detector in combination with a photo ionization detector (SRI 8610C Chassis). Turbid Sample Preparation. Turbid samples were prepared by introducing known masses of silica flour into pre-prepared ammonium nitrate or trichloroethylene aqueous samples in 30 ml clear glass vials, sealed with zero headspace using Teflon caps. These masses were varied to create solutions with a range of turbidity values, as measured in nephelometric turbidity units (NTUs). NTUs provide a standardized measure of the extent to which white light is scattered at an angle of 908 from the direction of the incident beam by particles suspended in a liquid relative to the same effect observed in a standard solution containing the polymer reaction byproduct of hydrazine sulfate and hexamethylenetetramine, in accordance with Environmental Protection Agency Method 180.1, and are a standard measure of water quality in environmental science. Samples were prepared with turbidity values ranging between 0 and 250 NTUs. While turbidity encountered in the natural environment tends to range from 0 to 50 NTUs,48 higher values were also explored to assess the consistency of observed trends and better model observed behavior. This range of turbidity values corresponds to 25–250 parts per million of turbidinducing medium by mass, or less than 95 parts per million by volume, of the studied sample, yielding a number density of 200 000 to 46 000 000 particles per cm3, for these very fine particles. All solutions contained particle sizes with terminal velocities long enough to remain stable for spectroscopic analysis, based on Stokes’ Law. Stokes’ Law characterizes the rate at which particles (typically assumed to be spheres) fall through a fluid by examining the forces acting on the particles attributable to gravitational attraction, buoyancy, and viscous drag, as follows: Vt ¼

d 2 ðqparticle  qfluid Þg 18g

ð1Þ

where Vt = terminal velocity in m/s, d = particle diameter in meters, qparticle = particle density in kg/m3, qfluid = particle density in kg/m3, g = acceleration due to gravity in m/s2, and g = fluid viscosity in kg/(m s). Over the grain size range explored, terminal velocities span approximately 0.1 to 1.4 mm/min, implying that tests should be completed within 15 min (for 5 lm particles) to 145 min (for 1.6 lm particles) of shaking to ensure suspension of the desired particles during testing at the mid-height of standard cuvettes. With the above in mind, all mixtures were manually shaken prior to testing and tested within allowable timeframes (typically within 6 min), with all tests

FIG. 1. 20 lJ Raman system schematic.

commencing at approximately 30 s after shaking. Note, too, that all test durations were well below the limits that might induce photochemical effects. Turbidity measurement was done with a HACH 2100P Turbidimeter, which makes use of both a 908 and transmitted light detector to ensure measurement stability. Samples were prepared in a 3.5 mL cuvette (3-Q-10-GL14C Starna Cells Inc.) for spectroscopic analysis. Raman System Overview. Two distinct Raman systems were employed in this study. The testing carried out with the ammonium nitrate samples made use of a closed path 532 nm 3 lJ pulsed laser (600 kHz rep rate, ,0.9 ns pulse duration, Teem photonics SNG-03E-000) with a fast charge-coupled device (CCD) detector. The 3 lJ pulsed laser is coupled into a 532 nm Raman probe (Inphotonics) using an anti-reflection coated mirror, 103 objective lens, and a multimode fiber coupler. The probe has a 105 lm excitation fiber and 200 lm collection fiber, each with a numerical aperture (NA) of 0.22. The probe is positioned at a working distance of 7.5 mm from the inside wall of the sample cuvette producing a spot size of approximately 160 lm diameter at this interface and has a subtended solid angle of collection of 0.30 sr. Note that the cuvette is secured in a spring-loaded holder against a fixed stopblock, enabling highly reproducible positioning. Scattered light is collected in a 1808 backscatter configuration back into the fiber probe. The collection fiber is then coupled directly to a CCD spectrometer (B&W Tek Prime X). Raman signal intensity at any given wavelength was obtained as the output of the 1024 3 58 CCD array as observed by the BWSpec software for a total integration time of 120 s. Tests involving TCE, which has a smaller Raman cross section than ammonium nitrate, required greater laser power and were therefore performed with an open-path 532 nm 20 lJ pulsed laser (2 kHz repetition rate; 0.4 ns pulse duration, Teem photonics PNG-002025-100). Here the 20 lJ excitation beam (1) is directed to an Ø 25.4 mm dichroic laser beam filter (Edmund Optics) at 458 angle of incidence (2), which directs the excitation energy into an optical train that focuses the energy on a sample cuvette (3). Within the optical train an Ø 45 mm anti-reflective (AR) aspheric lens (4) with a 0.612 NA (Thorlabs) focuses the excitation on the sample and serves as a collection and focus optic for the 1808 backscattered Raman photons from the sample. This configuration produces a spot size of approximately 0.23 mm diameter at the inside wall of the cuvette at a working distance of 35.6 mm and leads to a subtended solid angle of collection of 0.83 sr. Again

FIG. 2. Raman spectrum of a 2000 ppm aqueous nitrate sample revealing investigated nitrate (1067 cm1) and water Raman vibrational modes (3393 cm1).

here, the cuvette is positioned in a spring-loaded holder against a fixed stop-block to ensure reproducible positioning. Collected energy is directed through the dichroic filter (2) to an Ø 50.8 mm AR achromatic doublet lens (5) (Thorlabs) with NA of 0.17. The achromatic lens in turn focuses Raman photons on the entrance slit of a monochromator (7) (Oriel 130 1/8 m, with flat-ruled 1200 lines/mm grating blazed at 500 nm). A 532 nm Ø 25.4 mm long-pass filter is placed at the monochromator entrance slit (6) to eliminate source wavelength background. Energy passing directly through the dichroic filter is focused through a convex lens (8) and used to create a data acquisition trigger via a photodiode (9). Light passing through the monochromator is observed using a photomultiplier tube (PMT) (Hamamatsu H7422-40P) (10) operated in a photon-counting mode. The PMT is observed via an impedance matched link to a pre-amplifier (Ortec 9326, operated at a gain of 203) (11) and a 100 ps binned comparator (Ortec 9353) that enables repeated timeresolved data acquisition (12). A schematic of this system is shown in Fig. 1. For all experiments performed with this system, spectra were collected with 25 lm slits, a step size of 0.1 nm, and 120 s observation time per wavelength, and integrated for the duration of each pulse of the excitation laser.

RESULTS AND DISCUSSION Results from Ammonium Nitrate Investigation. All turbidity-related Raman analyses involving ammonium nitrate focused on observing the NO3 symmetric stretching peak at 1067 cm1 and the intermolecular OH stretching mode of water at 3393 cm1 (see Fig. 2). Three aqueous concentrations of ammonium nitrate were studied: 2000, 3500, and 6800 ppm nitrate-N under varying turbidity levels. While these concentrations are generally higher than those found in the environment (researchers have reported field concentrations as high as 2000 ppm,49,50) their use facilitated improved signal to noise in the analyses allowing observation of even subtle trends in the relationship between sample turbidity level and Raman signal strength. Results of the Raman analyses of the silica doped ammonium nitrate solutions are presented in Fig. 3. Open symbols represent turbid solutions, and the solid

APPLIED SPECTROSCOPY

1385

FIG. 3. Intensity of ammonium nitrate Raman line (1067 cm1) as a function of turbidity (NTU) for varying aqueous nitrate solution concentrations.

symbols represent average intensities in zero turbidity, or reference, solutions developed through a calibration exercise. The spread in the data tends to increase at higher nitrate concentrations as would be anticipated due to the fact that variability in Raman intensity typically scales with the square root of absolute intensity. For all three studied nitrate concentrations, the intensity of the NO3 Raman return tends to drop rapidly over the fieldrelevant NTU range (60–50 NTUs) and then decline more gradually at higher NTU values. A similar trend is noticed in the intensity of the OH water stretching Raman return at 3393 cm1, as shown in Fig. 4. Overall, the presence of suspended fine silica flour with grain sizes in the range 1.6–5.0 lm has a detrimental impact on the intensity of the Raman observations. However, unlike the NO3 signal intensity which varied in magnitude in relation to the ammonium nitrate concentrations, the nitrate concentration in solution has no discernible impact on the observed variation in the intensity of the OH stretching line at 3393 cm1 over the studied nitrate concentration range. At zero turbidity, the water line intensity remained consistent within 61.5%.

FIG. 4. Raman intensity of the OH stretching line (3393 cm1) of water in ammonium nitrate aqueous solution as a function of turbidity (NTU).

1386

Volume 68, Number 12, 2014

FIG. 5. Nitrate Raman line intensity as a function of solution turbidity with turbidity-inducing silica grain size as a parameter in 2000, 3500, and 6800 ppm nitrate solutions.

This observation can be attributed to the fact that the Raman vibrational modes of the water are unaffected by the presence of the nitrate at these levels. The observed yet insignificant variation at zero turbidity is likely attributable to expected fluctuations in Raman line intensity, which are known to be on the order of the square root of the number of Raman scattered photons, as well as competitive partitioning of the excitation source energy between nitrate and water molecules in solution, particularly at the higher nitrate concentrations. Initial review of the observed trends relating Raman signal intensity and turbidity suggests that there is potential to predictably correct for the influence of turbidity. However, before pursuing that avenue, it is important to first determine if turbidity level is indeed the driving factor underlying these observations. Another potentially influential variable is the grain size of the suspended particles inducing the solution turbidity, which is investigated in the following section. Influence of Suspended Particle Size on Raman Monitoring of Ammonium Nitrate (NH4NO3). In order to assess the influence of suspended silica flour grain size on Raman measurements, tests were performed on aqueous ammonium nitrate solutions at varying levels of turbidity, each created with differing levels of ammonium nitrate (again, 2000, 3500, and 6800 ppm) and various ranges of suspended particle grain size (1.6, 2.1, 2.4, 3.5, and 5.0 lm). Figure 5 displays the intensity of the Raman nitrate line as a function of solution turbidity with silica grain size as a parameter for the 2000, 3500, and 6800 ppm nitrate solution concentrations. In addition, the inset in the upper right corner of the figure highlights the relationship between observed turbidity levels and the silica content in the tested solutions (g/ml). From Fig. 5, it is clear that the intensity of the nitrate Raman return decreases with increasing turbidity for all three ammonium nitrate concentrations studied as noted earlier (this same trend was highlighted earlier in Fig. 3). However, it also shows that suspended particle size has no perceptible effect on Raman signal

FIG. 6. Normalized Raman intensity of the OH stretching line (3393 cm1) of water and ammonium nitrate Raman line (1067 cm1) in ammonium nitrate aqueous solution as a function of turbidity (NTU).

intensity, as the intensities of the nitrate Raman return span a common range for each of the studied nitrate concentrations at any given level of turbidity regardless of the turbidity-inducing grain size. Note, too, that with the highlighted relationship between turbidity and silica content as a function of grain size shown in the figure inset, attempts were also made to relate the Raman intensities observed at different turbidity levels to turbidity inducing silica mass per unit volume in the tested solutions as well as implied particle surface area (based on a spherical particle assumption and knowledge of particle-specific gravity (2.65)). However, these perspectives had no discernible benefit over the relationship to turbidity already presented. And, of course, assessing mass or surface area of suspended particles in an actual in-field situation is impractical, while determination of turbidity is quite feasible even in single-sided test geometries as commonly pursued in field work. A similar set of analyses (not shown) for the water Raman return also shows that Raman intensity is driven by overall turbidity level and has no apparent correlation to the size of the suspended particles that induce that turbidity within the studied grain size range. Beyond these observations, statistical analyses were pursued to provide quantitative assessment of any potential relationship between Raman return intensity from a turbid solution and the size of the suspended particles in the solution. First, a one-way general linear model ANOVA analysis was performed on the entire dataset using Minitab, version 16. The normalized intensity of the nitrate Raman line was used as the response to changes in the studied variables, suspended particle grain size, and turbidity level, respectively. The probability, P, that turbidity level does not affect the Raman nitrate signal was 0.000, whereas the P value for suspended particle grain size was 0.037, with a = 0.05, indicating that grain size has no significant effect. An additional twoway balanced ANOVA analysis was also run by

extracting a balanced data set from the overall sample at nine narrow turbidity ranges (63 NTUs) for which three response observations were present for each of the five studied grain sizes. Here the probability, P, that turbidity level does not affect the Raman nitrate signal was again 0.000, whereas the P value for suspended particle grain size was 0.448, with a = 0.05, reinforcing that at any given level of turbidity, grain size likely had no influence on nitrate Raman intensity. From the above analysis it may be concluded that at any given level of turbidity, the grain size of the silica driving the turbidity level has no significant effect on the intensity Raman observations, across the studied grain size range 1.6–5.0 lm. This result likely stems from the narrow range of particle sizes explored here as well as the absolute size and morphology of the particles investigated which are angular in shape and have a size parameter (pd/k) from  9 (1.6 lm) to 30 (5.0 lm), given the employed 532 nm excitation wavelength. In this context, since turbidity level, s, is proportional to the obscured cross section of the studied sample as defined by the relation s = N R, where N represents the number of suspended particles in the medium and R represents the scattering cross section of a suspended particle,51 as the particle size increases, the number of particles required to achieve any given level of turbidity decreases (and vice versa), so that the total scattering cross section of the overall sample remains approximately the same regardless of the particles used to create it. Further, given the 1808 backscatter configuration of the employed system, and the well-established tendency for particles larger than the wavelength of light to induce much greater scattering in the forward rather than backward direction, any changes in the particle sizes employed to achieve a given level of turbidity are likely to influence only a small fraction of the backscattered light and thus will have a limited impact on the net effect on Raman observations collected via the employed 1808 backscatter optical configuration. Turbidity Correction for Turbid Ammonium Nitrate Samples. Based on the above analysis, the influence of turbidity on the intensity of Raman returns in the studied solutions can be modeled as a function solely of turbidity level. With this in mind, Fig. 6 presents all of the turbid solution nitrate and water Raman line data from all studied nitrate solution concentrations, normalized by their respective zero-turbidity intensities. The overall data set falls into a band in which there is a rapid rate of decline in Raman intensity at low turbidity values and a slowing rate of decline at higher turbidity levels, which is thus best fit with a negative natural logarithmic function relating normalized Raman line intensity to turbidity level as shown in the figure. Based on theoretical considerations, one might expect the relationship between turbidity level and Raman intensity to be exponential, as the Beer–Lambert law would predict attenuation of the incident excitation energy (and thus the directly proportional Raman return) to be governed by the factor (1  esl), where s is the turbidity level and l is the path length of the incident energy through the sample.51 However, the Beer– Lambert law in its most general formulation does not

APPLIED SPECTROSCOPY

1387

account for the presence of scatterers. While this can be accommodated through use of a combined extinction coefficient (here approximated by s) that accounts for both scattering and absorption, even then there is a presumption that the scatterers in the medium meet Rayleigh (pd/k , 1) or Mie (pd/k  1) criteria, depending on the formulation.52–55 As noted earlier, this is not necessarily the case here given the non-spherical shape and size parameters of the suspended particles in the tested samples. Further, the modified Beer–Lambert law also presumes that scatterers, if present, exist at low enough volume fractions that they do not induce multiple scattering phenomena. This is also unlikely to be the case in the tested samples, particularly at higher turbidity levels. Finally, the presence of the non-target suspended particles in samples tends to reduce the volume of the liquid medium that is interrogated by the incident energy, leading to a decline in observed scattering intensity from target analytes, relative to an unobstructed test volume.2,42,55 The net effect of these influences is that the backscattered Raman return of the target analyte tends to decrease more rapidly with increasing turbidity level than would be predicted by a simple negative exponential relationship to turbidity level. This observation is consistent with the findings of several research studies in adjacent fields. Zardecki,56 in an exploration of the limits of the Beer–Lambert law, highlighted that when the size of the scattering particles in a medium is greater than or comparable to the incident wavelength, the typical exponential loss of optical transmission through a turbid fluid must be multiplied by a correction factor which is a non-linear function of optical depth (a quantity proportional to turbidity) to account for increased forward scattering, thus reducing backscattered return. Numerical modeling work on aerosols by Wind and Szymanski57 and experimental and modeling efforts by Berrocal et al.58 on homogeneous turbid solutions of mono-disperse polystyrene spheres in distilled water reinforce the notion that multiple scattering tends to increase the light scattered in the forward direction in a system and drives a related divergence from the Beer–Lambert calculation. Further, Yang et al.59 studied the effect of the addition of an intralipid as a scatterer into mixtures of acetone, ethanol, and ethyl acetate and showed that the presence of the scatterers led to a non-linear relationship between target concentration and target Raman intensity due to multiplicative effects associated with the presence of the scatterers and related variations in the intensity of the excitation radiation reaching the target analyte and the actual sample volume being analyzed. Finally, and perhaps most closely related to the work described herein, Wu et al.60 recently reported a natural logarithmic relation between the backscatter coefficient of natural lake water and the concentration of suspended particulate inorganic matter in the water. With the above outlined relationship between turbidity and normalized Raman intensity in hand, a correction to manage turbidity effects on in situ analyte Raman measurements can be inferred. For the simple nitrate solution here, the inferential determination of in situ

1388

Volume 68, Number 12, 2014

nitrate concentration can be initiated by first comparing the observed in situ intensity of the water Raman line with a non-turbid standard. Then, because the relationship outlined in Fig. 6 implies that the water line intensity and nitrate line intensity are affected in the same manner by turbidity, the relative change in the Raman intensity of the water line can be used to correct the nitrate line intensity to a zero-turbidity equivalent (and, if desired, to infer the likely turbidity of the tested solution). The corrected measurement of the nitrate line intensity can then be correlated to a standard calibration curve relating the Raman intensity of the nitrate line to the concentration of ammonium nitrate in solution to approximate the in situ nitrate concentration. In the event that turbidity does not affect the target analyte and water Raman returns in the same manner, for example, due to non-linear response characteristics of the employed hardware, or varying absorption or scattering properties of the test medium at different Raman shifts, the ratio of the in situ to zero-turbidity water line intensities could be used to infer in situ turbidity and then an experimentally derived curve specifically relating normalized target analyte Raman return to turbidity could be utilized to obtain a corrected analyte measurement. To assess the effectiveness of this corrective approach, a series of validation tests were conducted on turbid samples with ‘‘unknown’’ turbidity values and varying concentrations of ammonium nitrate. Two concentrations were studied, 800 and 1450 ppm. In this particular case, because nitrate and water share the same curve relating normalized intensity to turbidity values, it is possible to simply normalize nitrate line intensities by the water line intensity obtained in respective tests. In this case, the error (represented by the standard deviation of the corrected observations relative to the actual concentration) between observed and actual concentrations is 4.0% and 2.2%, respectively, for the 800 and 1450 ppm data sets. In some cases however, the normalized intensity versus turbidity curves for nitrate and water may not be the same, and it will thus be important to employ the normalized water intensity versus turbidity relationship to infer turbidity, and then use the nitrate normalized intensity versus turbidity curve to estimate a correction factor for the nitrate observations. This approach is illustrated here by utilizing the normalized water intensity from each of the validation tests to obtain an estimated turbidity value that could be associated with each nitrate observation. Figure 7 presents the relationship between predicted and actual nitrate concentration at the studied turbidity values, for both the 800 and 1450 ppm data sets. When employing the normalized correction curves, there is error in the estimate of actual turbidity. This error is small at field relevant turbidity values, but increases significantly at higher turbidity values due to the shape of the negative natural logarithmic curve highlighted in Fig. 6, which flattens out as turbidity increases. However, this same flatness of the relationship between turbidity and normalized intensity also limits the error when turbidity is ultimately translated back into a correction

FIG. 7. Relationship between predicted and actual nitrate concentration in turbid aqueous solutions based on turbidity correction.

factor for intensity. As a result, the error in the corrected concentrations relative to actual for the 800 and 1450 ppm data sets, respectively, is a maximum of 9.2% and þ4.8%. As outlined above, the corrective procedure described here can likely serve as an effective means to determine actual concentration of compounds (here ammonium nitrate) in a turbid influenced in situ measurement scenario. Turbidity Correction for Turbid Chlorinated Solvent Samples. Discussion here focuses on interpretation of the effects of turbidity on time-resolved Raman spectroscopic (TRRS) measurements of TCE. Since the analyses of ammonium nitrate solutions described above revealed that variation in turbidity-inducing particle size has no significant effect on Raman observations in turbid samples for the studied grain size range, experiments here simply targeted development of a relationship between turbidity level and the intensity of the 381 cm1 (dCH) Raman shift of TCE for turbid aqueous solutions of TCE created using silica with a grain size of 5 lm. TRRS measurements were performed using the 20 lJ open path laser configuration described earlier which makes use of a photomultiplier tube for photon detection, and thus observed intensities

FIG. 8. Intensity of the dCH TCE Raman line (381 cm1) as a function of turbidity (NTU) for varying concentrations of TCE in aqueous solution.

are reported in counts. Tests were performed on samples at varying TCE concentrations (300, 600, 900, 1200 ppm) and turbidity levels. Results, shown in Fig. 8, are similar in trend to the prior investigation of ammonium nitrate and again reveal a rapid reduction in Raman counts as turbidity levels increase in the field-relevant range, followed by a plateau in the effect at higher turbidity levels. After normalizing each concentration-specific data set by its respective zero-turbidity TCE Raman line intensity (see Fig. 9), it is clear that the observed reduction in counts with increasing turbidity can again be described by a negative natural logarithmic function. However, the coefficients and constants in the equation are notably different from those obtained for the nitrate analysis. The difference stems from the need to employ a more powerful laser source (20 lJ versus 3 lJ) and more sensitive data acquisition system (PMT versus CCD) to work with TCE, which leads to a fundamentally different transfer function for the system. This implies that an equipment specific calibration will be required to interpret the effect of turbidity on any particular target analyte.

FIG. 9. Normalized intensity of the dCH TCE Raman line (381 cm1) as a function of turbidity (NTU) for varying concentrations of TCE in aqueous solution.

APPLIED SPECTROSCOPY

1389

Although at any given TCE concentration turbidity level can be expected to influence the Raman return of the water line in the same manner as the TCE Raman line, the water Raman line cannot be used as an intrinsic calibration in this case due to the interactions that are known to occur between the hydrogen bonds in water and the chlorine comprising the chlorinated solvent.46 Thus, an alternative mechanism to determine in situ turbidity would be required to identify the appropriate correction factor from Fig. 9 that could be employed to adjust field measurements. To this end, several researchers41,61,62 have successfully employed observations of diffuse reflectance to interpret the turbidity level of turbid media. Measured diffuse reflectance spectra contain information describing the scattering and absorption of light in a particular turbid medium, and the intensity of the diffuse reflectance tends to be inversely related to the turbidity of the sampled media. In addition, there are a range of singlesided turbidity meters available that could readily be employed in situ to acquire direct measures of the turbidity level.63,64 Direct turbidity observations, or inferred levels of turbidity developed from diffuse reflectance observations, combined with a turbidity influence curve such as that shown in Fig. 9, thus offer means to initiate the correction of Raman data obtained in the presence of analytes that alter the Raman response of water.

CONCLUDING REMARKS Turbidity has been shown to adversely influence the intensity of Raman spectroscopic observations of nitrate and TCE in aqueous solution, with a consistent trend of natural logarithmic decreases in Raman return with linear increases in turbidity. Physically, this highlights that the presence of the non-target-suspended particles in samples tends to alter the scattering properties of the test solution and reduce the volume of the liquid medium that is interrogated by the incident energy, leading to a decline in observed scattering intensity from target analytes, relative to an unobstructed test volume. The form of the relationship is significant in that it highlights that turbidity induces significant changes in Raman return intensity at field relevant turbidity values. This trend appears consistent regardless of turbidity-inducing grain size over a tested grain size range of 1.6 to 5.0 lm. A correction approach relating water Raman signatures to turbidity level and, in turn, to the influence of turbidity on analyte signatures has been described to adjust turbid sample Raman measurements for the presence of suspended solids in situations in which the analyte and/or sample background do not influence the vibrational modes of water. In this study, this approach provided reliable indications of actual liquid phase analyte concentrations in turbid solutions with less than 10% error. Turbid sample Raman measurements for systems containing an analyte that does influence the vibrational modes of water still appear to follow a predictable reduction in intensity as a function of turbidity and could potentially be corrected by employing

1390

Volume 68, Number 12, 2014

direct turbidity observations or diffuse reflectance measurements. Water was chosen in this work as a high volume fraction component of the studied solutions that would have a reference Raman signature that could be known a priori in an actual field situation, yet would have a strong Raman signature that would still be susceptible to turbidity effects in the field, and thus indicative of turbidity influence. Conceivably, other solvents representing a significant fraction of the tested medium that (a) could be anticipated to be present in the target mixture, (b) would have a strong Raman response susceptible to turbidity effects, and (c) would be unaltered by the presence of other components of the test solution, could function in a similar manner, implying that this work may have applicability beyond solutions for which water is the principal solvent. In addition, consistency in the relationship relating turbidity level to changes in observed Raman intensities of both water and varying concentrations of a target analyte, as well as across solutions with equivalent turbidity levels created using varying suspended particle grain sizes, highlights that the turbidity correction, while system dependent, is likely analyte independent, and thus broadly applicable, for analytes that do not influence the Raman return of the medium in which they are dissolved. ACKNOWLEDGMENT This material is based upon work supported by the National Science Foundation under Grant No. 0927112. 1. J. Bloch, B. Johnson, N. Newbury, J. Germaine, H. Hemond, J. Sinfield. ‘‘Field Test of a Novel Microlaser-Based Probe for In-Situ Fluorescence Sensing of Soil Contaminants’’. Appl. Spectrosc. 1998. 52(10): 1299-1304. 2. J.V. Sinfield, J.T. Germaine, H.F. Hemond. ‘‘Effects of Soils on Laser Induced Fluorescence of BTX Contaminated Pore Waters’’. J. Geotech. Geoenviron. Eng. 1999. 125(12): 1072-1077. 3. M.L. Kram, A.A. Keller, J. Rossabi, L.G. Everett. ‘‘DNAPL Characterization Methods and Approaches, Part 1: Performance Comparisons’’. Ground Water Monit. Rev. 2001. 21(4): 109-123. 4. M.G. Serrato, R.S. Van Pelt, S.R. McMullin, J. Rossabi. ‘‘Application of Cone Penetrometer Testing Technology in Waste Site Investigations at SRS’’. Trans. Am. Nucl. Soc., ANS Annual Meeting, San Diego, CA, June 20–24, 1993. 68(A): 54-55. 5. J.M. Andrews, S.H. Lieberman. ‘‘Multispectral Fluorometric Sensor for Real Time In-Situ Detection of Marine Petroleum Spills’’. Proceedings of the International Conference on Oil and Hydrocarbon Spills, Modeling, Analysis and Control. Oil Spill 98. Southampton, UK; July 28, 1998. Pp. 291-301. doi: 10.2495/OIL980261. 6. J.D. Pasteris, B. Wopenka, J.J. Freeman, P.G. Brewer, S.N. White, E.T. Peltzer, G.E. Malby. ‘‘Raman Spectroscopy in the Deep Ocean: Successes and Challenges’’. Appl. Spectrosc. 2004. 58(7): 195A208A. 7. T.M. Battaglia, E.E. Dunn, M.D. Lilley, J. Holloway, B.K. Dable, B.J. Marquardt, K.S. Booksh. ‘‘Development of an in Situ Fiber Optic Raman System to Monitor Hydrothermal Vents’’. Analyst. 2004. 129(7): 602-606. 8. H.D. Kronfeldt, H. Schmidt, M. Maiwald, L.-H. Gallasch, J. KonatStepowicz, M. Lehaıtre, A. LeNoac’h, J. Pfannkuche, H. Amann, M. Szymczak-Zyla, A. Filipowska, L. Lubecki, G. Kowalewska, O. Esteban, M.-C. Naverrete, N. Dıaz-Herrera, A. Gonzalez-Cano, E. Bernabeu, C. Gibson, B. Mac Craith, M. Leclercq, B. Roussel. ‘‘Multiparametric In-Situ Spectroscopic, Measuring System for Coastal Monitoring Employed Under Field Conditions in the Gulf of Gdansk’’. In: T. Matsui, J.S. Chung, J.-L. Michel, H. Allersma, editors. Proceedings of the 14th International Offshore and Polar Engineering Conference, Toulon (France), May 23–28, 2004. 433437.

9. T.S. Moore, K.M. Mullaugh, P.R. Holyoke, A.S. Madison, M. Yucel, G.W. Luther. ‘‘Marine Chemical Technology and Sensors for Marine Waters: Potentials and Limits’’. Ann. Rev. Mar. Sci. 2009. 1: 91-115. 10. A.P. Ivanov, I.I. Kalinin, A.I. Kolesnik. ‘‘Application of Lasers in Oceanographic Investigations’’. J. Appl. Spectrosc. 1982. 37(4): 1097-1103. 11. D.M. Phillips. ‘‘The Effect of Water Turbidity on Laser Depth Sounding Performance’’. In: M.F. Penny, D.M. Phillips, editors. Proceedings, Fourth Laser Hydrography Symposium, Adelaide, Australia, September 30–October 3, 1980, 1981. Document AR-002576, Report ERL-0193-SD, March. 12. F.E. Hoge, R.N. Swift. ‘‘Airborne Detection of Oceanic Turbidity Cell Structure Using Depth-Resolved Laser-Induced Water Raman Backscatter’’. Appl. Opt. 1983. 22(23): 3778-3786. 13. H. Amann. ‘‘Laser Spectroscopy for Monitoring and Research in the Ocean’’. Phys. Scr. 1998. T78: 68-72. 14. B.R. Marshall, R.C. Smith. ‘‘Raman Scattering and in-Water Ocean Optical Properties’’. Appl. Opt. 1990. 29(1): 71-84. 15. P.-A. Schnegg. ‘‘An Inexpensive Field Fluorometer for Hydrogeological Tracer Test with Three Tracers and Turbidity Measurements’’. In: E. Bocanegra, D. Martinez, H. Massone, editors. Articles of the Geomagnetism Group at the University of Neuchaˆtel, Groundwater and Human Development. 2002. 1484-1488. 16. I.V. Vellekoop, C.M. Aegerter. ‘‘Scattered Light Fluorescence Microscopy; Imaging Through Turbid Layers’’. Opt. Lett. 2010. 35(8): 1245-1247. 17. M.V. Schulmerich, W.F. Finney, R.A. Fredricks, M.D. Morris. ‘‘Subsurface Raman Spectroscopy and Mapping Using a Globally Illuminated Non-Confocal Fiber-Optic Array Probe in the Presence of Raman Photon Migration’’. Appl. Spectrosc. 2006. 60(2): 109-114. 18. I. Barman, G.P. Singh, R.R. Dasari, M.S. Feld. ‘‘Turbidity-Corrected Raman Spectroscopy for Blood Detection’’. Anal. Chem. 2009. 81(11): 4233-4210. 19. P. Matousek, N. Everall, M. Towrie, A.W. Parker. ‘‘Depth Profiling in Diffusely Scattering Media Using Raman Spectroscopy and Picosecond Kerr Gating’’. Appl. Spectrosc. 2005. 59(2): 200-205. 20. P. Matousek. ‘‘Raman Signal Enhancement in Deep Spectroscopy of Turbid Media’’. Appl. Spectrosc. 2007. 61(8): 845-854. 21. C. Eliasson, M. Claybourn, P. Matousek. ‘‘Deep Subsurface Raman Spectroscopy of Turbid Media by a Defocused Collection System’’. Appl. Spectrosc. 2007. 61(10): 1123-1127. 22. G. Braunlich, G. Gamer, M.S. Petty. ‘‘Detection of Pollutants in Water by Raman Spectroscopy’’. Water Res. 1973. 7(11): 1643-1647. 23. J. Rossabi, B.D. Riha, J.W. Haas III, C.A. Eddy-Dilek, A.G. Lustin, M. Carrabba, W.K. Hyde, J. Bello. ‘‘Field Tests of a DNAPL Characterization System Using Cone Penetrometer-Based Raman Spectroscopy’’. Ground Water Monit. R. 2000. 20(4): 72-81. 24. I. Bogrekci, W.-S. Lee. ‘‘Design of a Portable Raman Sensor for Phosphorus Sensing in Soils’’. Presented at: ASABE Meeting, Paper Number 051040, ASABE Annual International Meeting, Tampa, Florida, July 17–20, 2005. 25. G. Langergraber, N. Fleischmann, F. Hofstaedter, A. Weingartner. ‘‘Monitoring of a Paper Mill Wastewater Treatment Plant Using UV/ VIS Spectroscopy’’. Water Sci Technol. 2004. 49(1): 9-14. 26. B.D. Downing, B.A. Pellerin, B.A. Bergamaschi, J.F. Saraceno, T.E.C. Kraus. ‘‘Seeing the Light: The Effects of Particles, Dissolved Materials, and Temperature on In Situ Measurements of DOM Fluorescence in Rivers and Streams’’. Limnol. Oceanogr-Meth. 2012. 10: 767-775. 27. A. Diaspro. Confocal and Two-Photon Microscopy: Foundations, Applications, and Advances. New York: Wiley-Liss, 2002. 580 pp. 28. N. Everall, T. Hahn, P. Matousek, A.W. Parker, M. Towrie. ‘‘Picosecond Time-Resolved Raman Spectroscopy of Solids: Capabilities and Limitations for Fluorescence Rejection and Influence of Diffuse Reflectance’’. Appl. Spectrosc. 2001. 55(12): 1701-1708. 29. M.V. Schulmerich, W.F. Finney, R.A. Fredricks, M.D. Morris. ‘‘Subsurface Raman Spectroscopy and Mapping Using a Globally Illuminated Non-Confocal Fiber-Optic Array Probe in the Presence of Raman Photon Migration’’. Appl. Spectrosc. 2005. 60(2): 109-114. 30. A. Corlu, R. Choe, T. Durduran, M.A. Rosen, M. Schweiger, S.R. Arridge, M.D. Schnall, A.G. Yodh. ‘‘Three-Dimensional In Vivo Fluorescence Diffuse Optical Tomography of Breast Cancer in Humans’’. Opt. Express. 2007. 15(11): 6696-6716. 31. Z. Yaqoob, D. Psaltis, M.S. Feld, C. Yang. ‘‘Optical Phase Conjugation for Turbidity Suppression in Biological Samples’’. Nat. Photonics. 2008. 2: 110-115.

32. S.M. Popoff, G. Lerosey, R. Carminati, M. Fink, A.C. Boccara, S. Gigan. ‘‘Measuring the Transmission Matrix in Optics: An Approach to the Study and Control of Light Propagation in Disordered Media’’. Phys. Rev. Lett. 2010. 104(10): 100601.1-100601.4. 33. I. Barman, C.-R. Kong, N.C. Dingari, R.R. Dasari, M.S. Feld. ‘‘Development of Robust Calibration Models Using Support Vector Machines for Spectroscopic Monitoring of Blood Glucose’’. Anal. Chem. 2010. 82: 9719-9726. 34. J. Wu, M.S. Feld, R.P. Rava. ‘‘Analytical Model for Extracting Intrinsic Fluorescence in Turbid Media’’. Appl. Optics. 1993. 32(19): 3585-3595. 35. D. Oelkrug, E. Ostertag, R.W. Kessler. ‘‘Quantitative Raman Spectroscopy in Turbid Matter: Reflection or Transmission Mode?’’ Anal. Bioanal. Chem. 2013. 405: 3367-3379. 36. M.G. Muller, I. Georgakoudi, Q. Zhang, J. Wu, M.S. Feld. ‘‘Intrinsic Fluorescence Spectroscopy in Turbid Media: Disentangling Effects of Scattering and Absorption’’. Appl. Optics. 2001. 40(25): 4633-4646. 37. R. Weersink, M.S. Patterson, K. Diamond, S. Silver, N. Padgett. ‘‘Noninvasive Measurement of Fluorophore Concentration in Turbid Media with a Simple Fluorescence/Reflectance Ratio Technique’’. Appl. Optics. 2001. 40(34): 6389-6395. 38. Q. Zhang, M.G. Muller, J. Wu, M.S. Feld. ‘‘Turbidity-Free Fluorescence Spectroscopy of Biological Tissue’’. Optics Lett. 2000. 25(19): 1451-1453. 39. W.-C. Shih, K.L. Bechtel, M.S. Feld. ‘‘Intrinsic Raman Spectroscopy for Quantitative Biological Spectroscopy Part I: Theory and Simulations’’. Opt. Express. 2008. 16(17): 12726-12736. 40. K.L. Bechtel, W.-C. Shih, M.S. Feld. ‘‘Intrinsic Raman Spectroscopy for Quantitative Biological Spectroscopy Part II: Experimental Applications’’. Opt. Express. 2008. 16(17): 12737-12745. 41. C. Reble, I. Gersonde, S. Andree, H.J. Eichler, J. Helfmann. ‘‘Quantitative Raman Spectroscopy in Turbid Media’’. J. Biomed. Opt. 2010. 15(3). 42. W.-C. Shih. Quantitative Biological Raman Spectroscopy for Noninvasive Blood Analysis. [Ph.D. Dissertation]. Cambridge, MA: Massachusetts Institute of Technology, 2007. 43. T.L. Tipton, B.S. Vogt. ‘‘Self-Referencing Fiber-Optics Fluorescence Sensor for Turbid Samples’’. J. Environ. Eng.-ASCE. 1998. 124(6): 545-548. 44. M. Bristow, D. Nielsen, D. Bundy, R. Furtek. ‘‘Use of Water Raman Emission to Correct Airborne Laser Fluorosensor Data for Effects of Water Optical Attenuation’’. Appl. Optics. 1981. 20(17): 2889-2906. 45. C.K. Monwuba. Geoenvironmental Influences on Raman Spectroscopic Monitoring of Chlorinated Solvent Natural Attenuation. [Doctoral Dissertation]. West Lafayette, IN: Purdue University, 2013. 46. J.V. Sinfield, C.K. Monwuba. ‘‘Inferential Monitoring of Chlorinated Solvents Through Raman Spectroscopic Observation of the Vibrational Modes of Water’’. J. Contam. Hydrol. Paper submitted. 2013. 47. S.K. Parida, S. Dash, S. Patel, B.K. Mishra. ‘‘Adsorption of Organic Molecules on Silica Surface’’. Adv. Colloid Interface Sci. 2006. 121(1–3): 77-110. 48. C.J. Paul, R.W. Puls. ‘‘Impact of Turbidity on TCE and Degradation Products in Ground Water’’. Ground Water Monit. Rev. 1997. 17(1): 128-133. 49. S.M. Strathouse, G. Sposito, P.J. Sullivan, L.J. Lund. ‘‘Geologic Nitrogen: A Potential Geochemical Hazard in the San Joaquin Valley, California’’. J. Environ. Qual. 1980. 9(1): 54-60. 50. S.J. Smith, D.K. Cassel. ‘‘Estimating Nitrate Leaching in Soil Materials’’. In: R.F. Follett, D.R. Keeney, R.M. Cruse, editors. Managing Nitrogen for Groundwater Quality and Farm Profitability. Madison, WI: Soil Science Society of America, 1991. 165-188. 51. W. Heller, W. Pangonis. ‘‘Theoretical Investigations on the Light Scattering of Colloidal Spheres. I. The Specific Turbidity’’. J. Chem. Phys. 1957. 26: 498-506. 52. C.V. Raman. ‘‘On the Molecular Scattering of Light in Water and the Colour of the Sea’’. Proc. R. Soc. London A. 1922. 101(708): 64-80. 53. J.W. Ryde. ‘‘The Scattering of Light by Turbid Media. Part I. Proc. R’’. Soc. London. A. 1931. 131(817): 451-464. 54. D.W. Hahn. ‘‘Light Scattering Theory’’. University of Florida, Department of Mechanical and Aerospace Engineering, 2009. 13 pp. 55. M. Soos, M. Lattuada, J. Sefcik. ‘‘Interpretation of Light Scattering and Turbidity Measurements in Aggregated Systems: Effect of Intra-Cluster Multiple-Light Scattering’’. J. Phys. Chem. B. 2009. 113(45): 14962-14970.

APPLIED SPECTROSCOPY

1391

56. A. Zardecki. ‘‘Multiple-Scattering Corrections in the Beer-Lambert Law’’. Proc. SPIE 0410, Laser Beam Propagation in the Atmosphere. 1983. 103: 9 pp. doi:10.1117/12.935761. 57. L. Wind, W.W. Szymanski. ‘‘Quantification of Scattering Corrections to the Beer–Lambert Law for Transmittance Measurements in Turbid Media’’. Meas. Sci. Technol. 2002. 13(3): 270-275. 58. E. Berrocal, D.L. Sedarsky, M.E. Paciaroni, I.V. Meglinski, M.A. Linne. ‘‘Laser Light Scattering in Turbid Media: Part I: Experimental and Simulated Results for the Spatial Intensity Distribution’’. Opt. Express. 2007. 15(17): 10649-10665. 59. J. Yang, Z.-P. Chen, J. Zhang, J.-W. Jin, Y. Chen. ‘‘Quantitative Raman Spectrometry: The Accurate Determination of Analytes in Solution Phase of Turbid Media’’. Chemometr. Intell. Lab. 2013. 126: 6-10.

1392

Volume 68, Number 12, 2014

60. G. Wu, L. Cui, H. Duan, T. Fei, Y. Liu. ‘‘Absorption and Backscattering Coefficients and Their Relations to Water Constituents of Poyang Lake, China’’. Appl. Optics. 2011. 50(34): 6358-6368. 61. G. Zonios, A. Dimou. ‘‘Light Scattering Spectroscopy of Human Skin In Vivo’’. Opt. Express. 2008. 17(3): 1256-1267. 62. A.M. Enejder, T.W. Koo, J. Oh, M. Hunter, S. Sasic, M.S. Feld, G.L. Horowitz. ‘‘Blood Analysis by Raman Spectroscopy’’. Opt. Lett. 2002. 27(22): 2004-2006. 63. S. Mylvaganam, T. Jakobsen. ‘‘Turbidity Sensor for Underwater Applications: Sensor Design and System Performance with Calibration Results’’. In: Proceedings OCEANS ‘98, September 28–October 1, 1998. 1: 158-161. 64. A.-F. Omar, M.-Z. MatJafri. ‘‘Turbidimeter Design and Analysis: A Review on Optical Fiber Sensors for the Measurement of Water Turbidity’’. Sensors. 2009. 9(10): 8311-8335.