Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 33
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies Technologies for Chemical Measurements As part of its deliberations, the committee estimated the ease of implementing the chemical technologies described below for in situ applications (Table 3). MASS SPECTROMETRY (FENN ET AL., 1990) Mass spectrometry (MS) is useful for a variety of purposes in environmental monitoring and research, including characterization of proteins, distinguishing among sources of compounds by the ratios of various isotopes (e.g., 13C/12C), and measurement of heavy metals. MS is critical for reconstructing the pathways of carbon flow in marine systems. Construction of carbon budgets, recognition of controls on the carbon cycle, and calibration of a CO2 paleobarometer are also important in studies of Earth history. Success in all these studies depends critically on the development and wide availability of adequate MS instrumentation. Due to natural fractionations of stable isotopes, compounds produced in various groups of organisms have slightly different isotopic compositions. Individual compounds are "isotopically labeled" at their source. These natural labels can be followed through the marine carbon cycle, and effects of secondary processes can be dissected in detail. In the past, all those intercompound differences were invisible because the available methods required samples so large that preparative isolation of individual compounds for isotopic analysis was impractical. Recently, however, isotope ratio monitoring techniques have been developed. For analysis of individual organic compounds, the effluent of a
OCR for page 34
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies TABLE 3 Analytical Technology Applied to In Situ Measurements Chemical Sensing Sample Processing Data Analysis, Interpretation D Mass spectrometry S Chromatography/electrophoresis separations E Chemometrics E Electrochemistry E Communications E-S Fluorometry E Flow injection analysis/continuous flow analysis E Data storage/handling E Absorption spectroscopy D Raman spectroscopy D Robotics D Fourier transform infrared spectroscopy S Refractive index D Satellite D Piezoelectric mass sensing E-S Immunochemistry/biochemistry S Polymers/new materials D Recognition chemistry D = Difficult—Unlikely to see widespread application to in situ measurements in the next 15 years. S = Straightforward—With some investment, could see application in 5 to 15 years. E = Easy—No major technical barriers to application between now and 5 years hence. high-resolution gas chromatograph is routed to a microscale combustion furnace. In turn, the products of combustion (principally CO2, but also N2) are led to the ion source of an isotope mass spectrometer. As the bursts of gas from the combustion of individual chromatographic peaks pass through the ion source, the relative abundance of 13C or 15N can be measured with a precision of about 2 parts in 10,000, good enough to observe natural variations. The entire system operates under conditions of continuous flow, so hundreds of measurements can be made within a single chromatographic run. Determining the sources and fates of individual organic compounds will be a major theme in marine organic geochemical research in the future, and its dependence on MS instrumentation is extreme. Marine laboratories in other countries are quickly entering this area, but laboratories in the United States are experiencing difficulty in finding funding to purchase the necessary instrumentation. A mass spectrometer is a device in which a sample is ionized and the ions are sorted and counted on the basis of their mass-to-charge ratios. Ions can be formed by several methods. A recently developed method, in which an ion beam sputters a nebulized liquid vapor, leaves molecules with multiple negative or positive charges. This permits the identification of very large molecules. Also, solid samples of molecules with masses of up to 300,000 daltons can be vaporized from surfaces using lasers with wave
OCR for page 35
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies lengths matched to the absorption wavelengths of the molecules. Mass analysis can be performed on the resulting ions using a variety of techniques, including: electrostatic, magnetic, double focusing, quadrupole lens, ion trap, time of flight, and Fourier transform ion cyclotron resonance. These methods have potential relevance for marine chemistry, but must circumvent the extreme difficulties of measuring materials dissolved in seawater. MS can be used for characterizing dissolved organic material (DOM) in seawater and in surface films and particulate organic material (POM) and dissolved material associated with sediments. As organic matter can serve as binding sites for trace metals, DOM identity and concentrations must be known to understand trace metal kinetics in seawater. Characterization of oceanic DOM is also important for a better understanding of biological processes. At present, most of the dissolved organic compounds found in seawater are unidentified. Carbohydrates and proteins, measured as total (free plus combined) hydrolyzable amino acids and sugars, constitute 5 to 20% of the DOM. Other components, such as lipids and low molecular weight organic compounds (e.g., formaldehyde and pyruvate), occur at much lower concentrations. Another identifiable organic constituent of seawater is DMSP [dimethylsulfonio-propionate, a precursor of dimethyl sulfide (DMS)]. DMSP is an important source of DMS to the atmosphere, having a role in formation of cloud condensation nuclei. With the MS systems available now, several thousand different molecules could be identified in seawater in land-based laboratories. A remaining problem is to isolate these compounds, which are present at concentrations many orders of magnitude more dilute than the major dissolved constituents of seawater. Emphasis should be placed on solution of basic problems of collection and purification of substances dissolved in seawater. Many investigators are using MS to characterize the structures of organic compounds obtained from marine particulate organic matter from sediment traps and from suspended particulates. The origins and alteration of this material are of great interest, and the molecules within it are generally susceptible to structural analysis. There is a great need for advanced instrumentation in this area, particularly for MS-MS systems. American laboratories are generally falling behind those in other countries in equipping for these techniques. Combinations of chromatography and MS are useful for oceanography. Gas chromatography combined with MS is a mature technique for analyses of complex mixtures and isotopic analysis. Liquid chromatography can be combined with MS to examine polar, labile substances without chemical derivatization, and is already applied in a variety of marine laboratories. This technique is sensitive enough to detect analytes in the picomolar range. Another application of MS is in combination with capillary zone electro-
OCR for page 36
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies phoresis. Efficient chromatographic separation can be achieved with this combination. These methods are ready to use now for land-based applications, but are expensive. Oceanographers are aware of many MS methods, but because of the cost, such methods are not readily available. This problem could be solved by a time-sharing arrangement or the development of national facilities. Mass spectrometers are in general too bulky and require too much stability and power to be used on board ocean-going vessels. Quadrupole mass spectrometers are sometimes taken to sea, however, for various biological experiments, such as studies of respiratory gas exchange by marine organisms. With further advances in ion traps, more MS devices may be used on board ships. Unfortunately, the only way to determine analytes quantitatively is to calibrate the equipment with identical, isotopically labeled molecules. New developments, particularly using internal standards in chromatographic combinations, may alleviate this problem. Accelerator mass spectrometry (AMS) is now being applied to measure the age of carbon in seawater. AMS is revolutionizing the measurement of light isotopes such as 14C in dissolved CO2 as well as DOC in seawater, requiring as little as a few hundred milliliters of sample. The older beta decay counting methods required 150-liter samples that were difficult and expensive to obtain. An AMS facility began operating at the Woods Hole Oceanographic Institution in 1991 at a cost of several millions of dollars. Seawater samples collected by the WOCE program will be analyzed in this facility. This AMS facility may also eventually be used for measuring 10Be for marine geology research. Inductively coupled plasma-mass spectrometry (ICP-MS) is revolutionizing the measurements of refractory metals, such as titanium, and can provide a wealth of isotopic information that could only be obtained previously with great difficulty. ICP-MS has been used as a fast and sensitive technique for measuring 230Th in marine sediments (Shaw and Francis, 1991) and barium in seawater (Klinkhammer and Chan, 1990). For the future, advances in the capabilities of mass spectrometers can be expected (Table 4), developed by interdisciplinary groups of academic, government, and industry scientists. It is unlikely, though not impossible, that MS techniques will be appropriate for buoy development. ELECTROCHEMICAL TECHNIQUES This section focuses on techniques designed to probe the interaction of an electrode with a seawater sample, especially those that derive a signal from oxidation-reduction reactions. From the point of view of measurements important to chemical oceanography and ocean science, the following electrochemical techniques have the most likelihood of application in the
OCR for page 37
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies Table 4 Expected Advances in the Field of Mass Spectrometry 1991 1996 2006 Overall sensitivity of detection 1–100 picomoles 1–100 femtomoles Attomole Molecular Weight Laser matrix time of flight 300,000 daltons 106 daltons Electrostatic quadrupole 130,000 daltons 106 daltons Time of flight <500 daltons 10,000 daltons Electrostatic-sector >3000 daltons >10,000 daltons LC-MS Flow probe 5–50 picomoles Sector <1200 daltons ~5000 daltons Flow electrostatic quadrupole <1 picomole Femtomole Electrostatic-ion trap ~3000 daltons near future: potentiometry, constant-potential techniques at steady state, pulse voltammetry, stripping voltammetry, and coulometric titrations (see Whitfield and Jagner, 1981, for review). Of these, potentiometry employs the simplest instrumentation, in that it requires only an electrometer for measurement of the potential difference between the sensing electrode and a reference electrode. The other techniques additionally require a current source and control circuitry, usually packaged together as a potentiostat. Coulometric techniques also require a means to determine charge, which is often done by incorporating an electronic integrator into a potentiostat. The technology of modern electronics provides for these purposes robust and inexpensive instruments, many of which include microprocessors suitable for controlling the experiment and manipulating the resulting data. Potentiometry (Buck, 1984) The potential of an electrode is related directly to the activities, and thus indirectly to the concentrations, of the chemical species involved in the equilibria that establish the potential. The main virtues of potentiometry are simplicity, very low power requirements, and the possibly small size of the sensors. The main drawback is that an unwanted reaction may enter into determination of the potential and sensitivity may be poor. Potentiometric sensing, as a transducing technique, can be coupled with an infinite variety of novel chemical reactions to solve specific analytical problems, as described below in the section on new chemistry. A wide variety of potentiometric ion-selective electrodes (ISEs) already have been developed (see
OCR for page 38
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies TABLE 5 Potentiometric Ion-Selective Electrodes Glass H+, Na+ Solid State Silver sulfide-based pressed pellet type—Ag+, S2-, Cd2+, Cu2+, Pb2+, I-, Br-, Cl- Single crystal-F-(LaF3) Organic Membrane (Liquid or Polymer) Ion exchanger or charged carrier—NO3-, Cl-, Ca2+ Neutral carrier—K+, Ca2+, Li+, Na+, Mg2+, H+, CO32- Table 5), and several are now being used, or are potentially useful, for measuring key ocean elements. The most common use of direct potentiometry (as compared with potentiometric titrations) is for measurement of pH (Culberson, 1981). Most other cation electrodes are subject to some degree of interference from other major ions. Electrodes for sodium, potassium, calcium, and magnesium have been used successfully. Copper, cadmium, and lead electrodes in seawater have been tested, with variable success. Anion-selective electrodes for chloride, bromide, fluoride, sulfate, sulfide, and silver ions have also been tested but have not yet found wide application. New polymer membrane-based ISEs for nitrate and carbonate exhibit detection limits and selectivities that may be applicable for ocean measurements. In addition, a number of these ISEs can be used as internal transducers for the design of useful potentiometric gas sensors. For example, dissolved CO2 can be detected potentiometrically by using either a glass membrane electrode or a polymer-based carbonate ISE, in conjunction with an appropriate reference electrode, behind an outer gas permeable membrane. Novel differential pCO2 sensors based on two polymer membrane-type pH sensors have also been developed recently. Constant Potential Techniques at Steady State Electrochemical detection under convective conditions has been applied widely in freshwater measurements. In addition, seawater measurements have been combined with flow injection analysis (FIA) and high pressure liquid chromatography (HPLC) (D.C. Johnson et al., 1986). Well-developed commercial product lines exist, and detection limits are typically in the range of femtomoles. For in situ, shipboard, and land-based measurements employing HPLC or FIA, electrochemical detection could provide increased capabilities with respect to compound-specific detection and improved detection limits. The limitations of this approach are determined completely by the requirements of FIA or HPLC.
OCR for page 39
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies The use of microelectrodes for steady-state measurements is less well developed but has special promise for sensors and monitors in cases where power and size are important constraints (Montenegro et al., 1991). The size and shape of microelectrodes can be tailored specifically to the analytical problem. The nature of the electrode response depends on the time scale of the experiment. Electrode arrays can be fabricated with thousands of small elements. Lithographic procedures raise the possibility of very low per-item cost of manufacturing identical arrays. Pulse Voltammetry (Osteryoung, 1988; Osteryoung and O'Dea, 1986) Modern pulse voltammetry employs stepwise changes in potential, the sequence of which is controlled by software. Thus, the exact choice and timing of potential steps can be tailored to the specific analytical problem. Standard pulse sequences routinely employed in electroanalytical investigations include those of square wave, normal pulse, and other voltammetries. The current resulting from a chemical reduction is directly proportional to concentration and typically can be measured with a relative precision of 0.2% and an absolute accuracy of better than 1%. For typical conditions and electrodes of conventional size, the change in current per change in concentration is on the order of 70 nanoamperes per micromolar. Detection limits in the worst case are about 1 micromolar and are typically 0.1 micromolar; in favorable cases, nanomolar concentration detection limits can be achieved. The linear range, over which signal is proportional to concentration, is typically a factor of 104 to 106, for example 10 nanomolar to 1 micromolar. The response, though of low resolution, is specific to the reacting species. It is possible, therefore, to determine metals in a specific oxidation state [e.g., As(III)] or as a specific complexed ion. Particulate material does not affect the response of the electrodes to dissolved material. The electrochemical behavior of organic compounds tends to be similar for different compounds in which the same functional group is reacting. Thus, voltammetric measurements can be used to give semiempirical quantification of classes of compounds without going through the expense and difficulty of determining the identity of individual compounds present. Measurements can be carried out at characteristic times of 10 microseconds to 10 seconds, thus providing considerable scope for optimization. Stripping Voltammetry (Zirino, 1981; Shuman and Martin-Goldberg, 1984; Van der Berg, 1989) Voltammetry can also be employed in the stripping mode; that is, the material of interest is accumulated in or on the electrode, and once concen-
OCR for page 40
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies trated is voltammetrically ''stripped'' from the electrode. Concentration factors of 105 can be achieved routinely, and thus detection limits in the range 10-11 to 10-12 molar can be achieved. At these levels of concentration, fidelity of the sample becomes the factor that controls the quality of the result. This technique has been used widely in oceanographic science, especially for the determination of metals such as lead, copper, cadmium, and zinc. Stripping voltammetry at mercury electrodes yields detection limits in aqueous solution of 30 picomolar for copper and 0.3 nanomolar for zinc (Stoeppler, 1991). Problems of interferences by other elements in either aqueous or amalgam solution have been studied exhaustively, and well-documented procedures exist for eliminating these interferences. Stripping voltammetry has been used to study the distribution of Zn(II) between labile and inert complexes in seawater with total zinc concentration of 50 nM (Muller and Kester, 1990). Formation constants of Zn(II) with various inorganic complexing agents (e.g., Cl- and NO3-) have been measured by anodic stripping voltammetry at total zinc concentrations of 10 nM (Komorsky-Lovríc and Branica, 1987). Distributions between labile and inert complexes have been determined in samples with 15 nM Cu(II) and Zn(II) by anodic stripping at a micromercury electrode (Daniele et al., 1989). The technique of absorptive stripping voltammetry holds great promise for the determination of electroactive substances that can be adsorbed at an electrode. New developments in the use of microelectrodes for stripping voltammetry provide the possibility of extending the extensive body of laboratory procedures now employed to in situ measurements. Ligand competition voltammetry is also being carried out on seawater samples. Coulometry (Bard and Faulkner, 1980) Coulometry comprises a set of techniques in which the total charge required (not the current, as in potentiometry) to oxidize or reduce the chemical species of interest is measured. The prime virtue of coulometric techniques is that they link the quantity of substance determined directly to the quantity of electrical charge, and thus expensive and often difficult procedures for standardization or calibration can be minimized or eliminated. Coulometric procedures are robust theoretically in that Q = nFN, where Q is quantity of charge, n is the number of electrons in the reaction, F is the Faraday constant, and N is the number of moles of reactant. This simple relationship is little affected by environmental variations (e.g., temperature or ionic strength). Coulometric procedures are also robust technically, in that the charge is just the integral with respect to time of the current. Coulometry is also easy to automate and operate under remote control. The accuracy of coulometric determinations is typically as good as 0.1%. Preci-
OCR for page 41
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies sion of 0.01% can be achieved, and it may be possible to achieve accuracies this good in special cases. The coulometric principle can be applied in many different ways to determine specific substances. The most obvious and direct way is to electrolyze a sample of known volume at constant potential. However, coulometry can also be applied in titrations at constant current, employing a reagent that is oxidized or reduced in the process. Coulometry may also be employed in titrations by coulometric generation of reagent. Examples include the determination of CO2 by reaction with ethanolamine, with subsequent titration with coulometrically generated base. Another coulometric reagent generation is for determination of SO2 by coulometrically generated iodine. Coulometric titration is now the method of choice for determination of TCO2 in seawater (Johnson et al., 1987). Finally, it can be used for the generation of standards, particularly for substances that are difficult to prepare and store, and for in situ calibration. Coulometry may also be used for detection in liquid chromatography and flow injection analysis. New electrochemical techniques can be applied directly to presently employed amperometric methods to optimize the potential-time waveform and current sampling scheme. For example, steady-state amperometric measurements at constant potential may be converted to pulsed operation with synchronous sampling of current in a way that improves performance but is transparent to the user. Optimal sampling schemes for individual methods can be developed through laboratory research. This type of development is exemplified by the Endeco® pulsed oxygen sensor. For remote operation, as from buoys, improvement in performance of computers should make it possible to convert amperometric techniques to voltammetric techniques. Determining the voltammetric response (which might consist of 150 data points per concentration measurement) provides better statistical definition of the current-concentration relationship. Knowledge of the voltammetric response could also provide the ability to compensate directly for changes in the properties of the reaction being used to determine concentration. This approach is presently practical in a laboratory research setting and, assuming further improvements in technology of computers, should be possible in oceanographic applications in the near future. Again, this type of development could be made transparent to the user. The size, shape, and material of the electrode can be tailored to the application. Problems of response time, for example, often can be solved by using a smaller electrode. The use of new electrode materials, including possibilities for modification of the electrode surfaces, could lead to new measurement capabilities. This emphasizes the importance of the development of polymers, new materials, and recognition chemistry, as discussed in the section on new chemistry. Size, shape, and choice of material all present new technical opportunities that can be applied now, but also present many
OCR for page 42
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies possibilities for longer-term development. Routine practice in laboratories lags significantly behind present technical capabilities. Modern instrumentation is qualitatively superior to that of even the mid-1980s. Major commercial efforts have gone into the development of sensors based on microelectrodes fabricated by means of lithography. Commercial products have not resulted because markets have not been identified that would support the volume required for economical production. Deterioration of response with time for in situ measurements might be decreased by employing multiple electrodes and using each only once. For example, present lithographic technology can allow production of line electrodes 20 micrometers wide, individually addressable, with 80-micrometer spacing. A 10-centimeter-long strip would contain 1000 electrodes. The electrode material also may be manipulated to achieve more specific or more reliably controlled performance. Specific catalysts for desired reactions may be incorporated into the electrode material or bound to the surface of the electrode. A present example is the coating of a carbon or other inert electrode with a polymer film impregnated with a mercuric salt. The resulting electrode is catalytic for reduction of metals, such as Pb2+, that are soluble in mercury. This is an area of research that could pay off through qualitative improvements in accuracy, precision, and response time. Developments in electroanalytical chemistry are driven by technical advances in electronics, computers, and materials. Present scientific capabilities available in a research laboratory will be applicable for field measurements with the advent of smaller, less expensive, more powerful computers. Miniaturization of electrochemical cells, which can improve performance, especially response time, can be implemented most effectively in the context of miniaturization of control circuitry. Concomitant low cost could make disposable systems a practical reality. Sophisticated data analysis and data handling techniques can, with better facilities for computation, be handled in real time. SPECTROPHOTOMETRY Absorbance Spectrophotometry encompasses a broad family of techniques, but fundamentally is very simple: the analyte in a sample (perhaps with a reagent added) absorbs some fraction of light at some selected wavelength, which may be correlated with the analyte's concentration by Beer's Law. The essence of the method is to compare the amount of light absorbed by a sample in the presence and absence of a specific analyte. The instrumentation necessary for spectrophotometry has been available for nearly half a century; however, diode array detectors have changed the performance of
OCR for page 43
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies single-wavelength instruments radically. The various forms of spectrophotometer differ primarily in details of light sources, dispersive element, and detector. The classical Beckman and Cary spectrophotometers can measure absorbance in the ultraviolet (UV) to near-infrared (near-IR) (250 to 1500 nanometers). Classical IR spectrophotometers use different sources and bolometric detectors to cover the range from about 2500 to more than 10,000 nanometers. Fourier transform IR (FT-IR) spectrophotometers use a more complex technique to cover the same wavelength range, with significant advantages (see section on infrared below). Atomic absorption spectrophotometers make use of lamps producing line spectra of particular elements to detect volatilized atoms of those elements by their characteristic absorbance lines. Photoacoustic and thermal lensing spectrophotometries make use of the conversion of absorbed light into heat to detect absorbance indirectly by measuring sound evolved and deflection by changes in refractive index, respectively. For the purposes of chemical oceanography, spectrophotometry is limited to wavelengths from the UV to near-IR (250 to 1500 nanometers). The various forms of spectrophotometer differ primarily in combinations of light sources, dispersive elements, and detectors. The principle of spectrophotometric measurement involves the comparison of two similar light levels, leading to an inherent weakness in the technique. In particular, as the concentration of absorbing analyte decreases, the two light levels become more and more similar until they cannot be distinguished. Thus, the sensitivity and accuracy of the method are very dependent upon the stability of the light source, the precision of the detectors, and the reproducibility of the sensor matrix. As a rule of thumb, with non-solid state sources and detectors, absorbance differences of 0.0001 unit are difficult to measure. The maximum extinction coefficient for a molecule in the visible or near-IR region (determined by the oscillator strength) is about 2 × 105 liter mole-1 cm-1. Thus, the detection limit for classical spectrophotometry is perhaps a few nanomoles per liter of "solvent." More commonly, detection limits are set by the presence of interfering contaminants in the sample; this is especially true for complex media like seawater. Some spectrophotometric techniques work to enhance sensitivity or utility in other ways. The advent of semiconductor diode array detectors permits entire spectra to be acquired simultaneously instead of one wavelength band at a time. Also, automated spectrophotometric analyzers originally developed for clinical use have been adapted for use at sea when many samples must be analyzed over a period of time. Computational techniques for signal averaging, smoothing, integration, and data analysis have been widely implemented in modern instruments. From the standpoint of chemical oceanography, spectrophotometry is a mature technology that is widely used and accepted.
OCR for page 62
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies ions and molecules in seawater will be needed to advance the field of analysis of materials dissolved in seawater. Advances in chromatography in the past 20 years have been achieved both in the chromatographic columns and in detectors that measure concentrations of chemicals in the outflow from the column. Improvements in column chemistry and in flow and solvent characteristics have increased speed, selectivity, and resolution of chromatographic sewparations. Many of the spectroscopic techniques described in this report, including mass spectroscopy, are used at the effluent end of gas chromatographic or high-performance liquid chromatography columns to measure concentrations by a range of so-called ''hyphenated'' techniques such as GC-MS, GC-FTIR, HPLC-UV-fluorescence diode array detectors, and others. Recent developments in computer control of flow regimens and mixtures of carrier phase solvents show promise for improving chromatographic techniques. Capillary chromatography with supercritical liquids and field-flow fractionation also show promise (Pimental and Coonrod, 1987). The latter technique allows separation of larger macromolecules and particles by application of temperature differences or electric fields across a flow. Chromatographic systems are inexpensive, and are generally rugged enough to be operated on a research vessel or could be modified relatively easily for this purpose. It is possible to pass large volumes of water through an adsorbent in a device that is the size of an double-A battery, for preconcentration and separation. Ion chromatography is very promising for ocean measurements. Anions and cations with concentrations below parts per billion can be separated and detected. The chromatographic retention time provides information on the chemical form of the ion, its oxidation state, and whether it is complexed with another material. Future developments that may facilitate ocean measurements from vessels or buoys include miniaturization of chromatographic equipment (so less solvent is needed per analysis), new solvent transport systems, such as electrokinetic transport, to reduce power requirements on the pumps, and more sensitive detectors for liquid chromatography. Certain combinations of very short columns and flow injection analysis are also promising for real-time studies. Electrophoresis has been used for decades to separate organic molecules on the basis of molecular charge. Capillary zone electrophoresis combines high voltage with electrophoretic mobility to measure ionic compounds. By adjusting the pH, many organic compounds can be made ionic. If one uses fused silica capillaries, there is bulk (electro-osmotic) flow of solution. The combination of electrophoretic and electro-osmotic flows brings improved separation. Detectors are placed at the end of the flow, as in chromatography. This method can be used to determine the major ions in
OCR for page 63
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies seawater. Detection with on-column preconcentration is possible for concentrations down to 1 part per billion. There are hybrid methods that allow separation of neutral solutes by interactions similar to liquid chromatography. The power requirements for capillary zone electrophoresis are low, so it could be used to drive flow injection analysis. The minimum cost of these devices is $15,000 for commercial systems. Flow Injection Analysis and Continuous Flow Analysis Flow injection analysis (FIA) is a robust method for automating complex chemical analyses (Ruzicka and Hansen, 1988). It is relatively simple and can be adapted for use with a variety of detectors, including spectrophotometers, fluorometers, mass spectrometers, and electrochemical analyzers. It has been used on board ships to determine dissolved nutrients (Johnson et al., 1985) and trace metals (Sakamoto-Arnold and Johnson, 1987; Elrod et al., 1991). Unsegmented continuous flow analysis (CFA) systems based on the principles of FIA can operate in situ over the entire range of depths found in the ocean (Johnson et al., 1986a, 1989). An FIA system consists of a pump, injection valve, manifold of capillary tubing and a flow-through detector (Figure 4). The injection valve is omitted from CFA systems. Chemical reagents are mixed continuously by pumping them together with a carrier solution. The seawater sample is periodically introduced into the carrier stream by the injection valve in FIA or introduced continuously in CFA. The mixture then flows through the detector. Typical analysis times range from 1 to 8 minutes for chemical determinations in seawater. A variety of manipulations can be performed with FIA systems. These include concentration with resin columns, solvent extraction, dialysis, separation with membranes, and derivatization. FIA systems are commercially available and are also easily constructed. Flow injection analysis with chemiluminescence detection (FIA-CL) can be used to determine a variety of metals in seawater, including cobalt, copper, manganese, and iron, with detection limits in the picomolar concentration range. These reactions typically involve the formation of a metal complex with an organic ligand and oxidation of the complex. Reaction products in the excited state decay by emitting a photon of light which can be used to quantify the metal concentration. FIA-CL systems are capable of achieving precision and accuracy similar to those obtained by graphite furnace atomic absorption spectrophotometry systems. A second advantage of FIA and CFA is their ability to operate underwater for extended periods of time (Johnson et al., 1986b). It is often desirable to obtain high resolution profiles of dissolved chemicals to examine their spatial and temporal variability. CFA systems have been used to measure spatial variability in nutrient distributions in the upper ocean (Johnson
OCR for page 64
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies FIGURE 4 Schematic of the FIA system used to determine cobalt in seawater. This system incorporates two valves, which allows cobalt to be concentrated on a column of immobilized 8-hydroxyquinoline (8-HQ) and magnesium ions in the seawater matrix to be removed by washing the column with water at pH 5.6. Reprinted with permission from Sakamoto-Arnold and Johnson (1987). Copyright 1987 American Chemical Society.
OCR for page 65
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies et al., 1989) and in redox reactive chemical concentrations around hydrothermal vents (Johnson et al., 1986a). These systems have used conventional peristaltic pumps to propel the sample and reagents through the system. Good spatial and temporal resolution can be obtained with systems based on peristaltic pumps because they produce high flow rates. However, systems based on peristaltic pumps are limited to deployment times of no longer than a few days because of the lifetime of the tubing used in the pump. The feasibility of using osmotic pumps to propel the sample and reagents for longer deployments is now being explored. Osmotic pumps are widely used for in vivo drug delivery systems. They can run for extended periods of time and do not require an external power source, as they operate on the difference in the osmotic pressure between the external seawater and an internal filling solution (Theeuwes and Yum, 1976). Typical flow rates with osmotic pumps are 1 to 10 microliters per hour. A reagent pump would consume a reagent volume of about 10 milliliters per year, and a sample pump would accumulate a sample volume of about 100 milliliters per year. The response time of an osmotically pumped system would be limited by the time required to flush the flow cell of the detector. At these low flow rates, at least 12 minutes would be required to flush a 1-centimeter-path-length photometer cell. However, for long-term deployments of up to 1 year, sampling rates of once per hour would be sufficient. Osmotically pumped CFA systems are now operating in the laboratory, and ocean deployments are expected to begin in the near future. A CFA system based on osmotic pumps would be extremely simple, and quite inexpensive to build. Instruments based on this design begin to fall within the realm of chemical sensors. A sensor system is usually considered to be much less complex than the typical FIA system. However, the mechanical simplicity of an osmotically pumped CFA system would certainly fall within the framework of a chemical sensor (Ruzicka and Marshall, 1990). The primary constraints on development of FIA for oceanographic research do not now appear to be the mechanical and electrical systems needed to operate them on board ship or in situ. Long-term stability of reagents will be a problem for deployments with osmotically pumped systems. These reagents must be stable for year-long time periods. The stability problem might be solved by using the time release strategy employed in drug delivery. Robotics (Hurst and Mortimer, 1987) Two hundred and fifty million chemical analyses are performed each day in the United States for quality assurance, environmental monitoring, medical diagnostics, toxicology, forensics, commerce, research and development, and other purposes. This large number of analyses is due in part to
OCR for page 66
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies a rapid increase in the number of laws mandating chemical testing. The increase in testing has surpassed the increase in the training of analytical chemists, which creates the need for faster, high-quality automated analyses. The need for analyzing an increasing number of samples with a diminishing supply of trained analysts may be addressed by the development of automated chemical instrumentation. To promote the development and use of laboratory automation, the National Institute of Standards and Technology formed the Consortium on Automated Analytical Laboratory Systems in 1990, with members from industry and the federal government. Different analysis regimens, defined by operational complexity and the number of samples that must be run each day, require different types of automation (Figure 5). Dedicated automation includes processes that are highly automated for repetition of a single task. Although Technicon Autoanalyzers® have been taken to sea for 20 years, these fall into the dedicated automation category and are not good examples of the potential of flexible automation, which allows automation of numerous sequential steps. Flexible automation includes more sophisticated laboratory robots, with a spectrum of operational complexities and a moderate number of samples. These robots are called flexible because they can be reprogrammed for different tasks and for a series of tasks that comprise a laboratory procedure. In actual practice, however, once a robot system is built for a specific analysis, it is rarely reconfigured for another method. The use of robots does not always decrease expense. Software development often accounts for a large percentage of development costs. Setting up automated systems requires expertise in computers and mechanical engineering. Usually it takes so much time to remove system errors that robots are not very flexible. Automated processes should not necessarily be designed with the same steps as those used by human technicians, because of the different sensor sets. Robots have been used in laboratories for a number of applications, including ELISA, transfer of bacterial plaques from agar plates to liquid media, analysis of cranberry juice, and microwave dissolution of environmental samples. How does the performance of laboratory robots compare with that of technicians? Robots are usually slower than technicians because their sensor sets are inferior to those of humans and thus more testing is required to ascertain completion of actions. Robots do not get interrupted, and their performance does not degrade over long periods of time. Sampling rates by robots depend on the procedure being automated. Technicians usually do better than robots on accuracy and precision in the short run, but robots perform better over the long run. Robotic precision has been improved by switching from volumetric to gravimetric measurements. For example, in making chromatographic calibration standards, better than 1% precision can be achieved, using mixtures robotically prepared by gravimetry.
OCR for page 67
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies FIGURE 5 Decision matrix for choosing whether or not to automate a laboratory method, and if so, what kind of automation to use. Courtesy of Zymark Corporation, Hopkinton, Massachusetts, USA. Robotic errors tend to be major, so that errors can be spotted more easily. In fact, robot-based laboratory procedures can provide a record of all steps for quality control purposes. Because of vibrations, power stability, and particularly corrosion, commercial laboratory robotic systems available today would have problems on ships. This problem provides an opportunity for research engineers to develop means to modify some sections of the ship to improve power and platform stability. Environmental constraints of a sea-based system were never factored into the design of today's laboratory robots, but the systems could be modified somewhat to reduce these problems. For now, robots would be most feasible for land-based measurements. In the future, however, robots could be important for at-sea measurements, because continuous or repeated measurements are often made over the course of many days. The base cost for a robot is about $25,000. The average complete cost for an analysis system that actually does a routine analytical task is two to five times this amount. This includes the costs of set-up, creating custom devices, custom software, and training. Land-based systems are available now, and sea-based systems are a distinct future possibility. Time, talent, and money will be needed to develop them. Robots can be set up to carry
OCR for page 68
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies out virtually any kind of analysis. Sea-based robotic systems must be designed to handle the power problems, the vibration and stability problems, and the corrosion problems presented by a shipboard environment. Chemometrics Analytical chemistry deals with quantitative, numerical concepts and with data. Chemists use statistics routinely to analyze their data and to present it, but usually at an elementary level. A relatively new subfield within chemistry is chemometrics. Chemometrics involves the application of multivariate statistics, mathematics, and computational methods to chemical measurements. It exists at the interfaces among the larger fields of chemistry, mathematics and statistics, and computer science. The goals pursued are varied and include designing and selecting optimal measurement procedures and experiments, gathering the best-quality analytical data, and gleaning the maximum amount of useful chemical information from the data. Specific areas of focus include statistics, sampling, experimental design, optimization, signal processing, factor analysis, resolution, calibration, modeling and parameter estimation, structure-property relations, pattern recognition, library searching, and artificial intelligence. Statistical thinking and methodology pervade this list, although the emphasis varies among items. Martin et al. (1988) suggested that application of chemometric methods to data obtained from marine samples could help classify organic components in samples, detect patterns in the data that merit further study, and serve as a noise filter. The progress of the field of chemometrics can be traced with the aid of a series of six review articles published at 2-year intervals in Analytical Chemistry (for example, Brown, 1990) which cover noteworthy advances. A series of recent textbooks also cover the field (for example, Massart et al., 1988). Recent topical reviews on subjects related to multivariate analysis of analytical data have appeared (for example, Jurs, 1990). Relatively few formal courses in chemometrics exist at either the undergraduate or graduate level. Chemometrics education is largely ad hoc, via short courses, symposia, textbooks, and tutorials. A compilation of tutorials taken from Chemometrics and Intelligent Laboratory Systems is available (Massart et al., 1990). Topics included are experimental design and optimization, signal processing, multivariate calibrations, and model building. This brief discussion of chemometrics is meant to provide sources of details about chemometrics methods rather than giving the basic information itself. The books and review articles cited provide a thorough introduction to the major topics covered here that could be important for ocean measurements. Experimental design, response surfaces, and optimization of experiments are all important components of chemometrics. Many trade-offs exist among
OCR for page 69
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies the number of experiments performed, the number of variables that can be tested for effect, and the efficiency of approach to the optimum experiment with the fewest number of replications. The simplex method (Walters et al., 1991) has been employed widely for optimization, especially in chromatography. For simplex optimization, the number of experiments needed to reach an optimum set of conditions is less than all possible combinations. This method is evolutionary in that it directs the experiment toward the optimum, the computations are simple, and it is amenable to automation. Self-optimizing analytical systems based on simplex optimization have been described in the literature. The use of self-optimizing analytical instruments would be of great value for unattended instrumentation in ocean measurements. The application of standard experimental design methods to oceanic measurements should be a routine part of designing all experiments, particularly those that are expensive to perform and difficult to repeat. Calibration and mixture analysis addresses the methods for performing standard experiments with known samples and then using that information optimally to measure unknowns later. Classical least squares, iterative least squares, principal components analysis, and partial least squares have been compared for these tasks, and the trade-offs have been discussed (Haaland, 1992). These methods allow a superior spectroscopic experiment to be performed because they allow the calibration of the method to depend in an optimal way on the experimentation done. These calibration methods ensure that the best measurements are done. They are especially useful for noisy data with substantial interferences, such as those found in many types of ocean science measurements. Multivariate model building allows the construction of models to connect independent variables and observations. Multiple linear regression analysis, principal components regression, partial least squares, and many other advanced statistical techniques have been used. The goal of such investigations may be estimation of the parameters in a mathematical model to assess the relative importance of the independent variables on the dependent variable, or it may be to generate the ability to predict unknown observations. Validation of the models constructed using either internal validation (such as jackknifing or duplexing) or external prediction (true prediction of unknowns) is an important component of such studies. Another important component of such studies is the determination of how to choose the best set of independent variables from among those available. Principal components analysis, factor analysis, and partial least squares have been used successfully in this area, especially for spectral data. Classification and clustering methods allow the analysis of data when the observations to be explained are not quantitatively known. Discriminants can classify patterns of data into categories. Clustering methods can
OCR for page 70
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies seek common characteristics among subsets of data taken from a larger data set. These classes of analysis can be used fruitfully with very large data sets to search for patterns among the data. Certain chemometric methods have been applied to ocean science for some time, primarily for data analysis. Optimization and some other newer chemometric techniques are just beginning to find their way into marine chemistry. Often, it is desirable to put the data into categories or classes, and these pattern recognition methods are well suited to such tasks. Some of the software to support chemometric methods is available in commercial form. Other software can be located by searching the literature or compilations such as the Scientific Computing and Automation Resource Directory. However, many methods of chemometrics are available only from the developers or through reports in the scientific literature rather than as software. It is likely that software well suited to ocean science would have to be produced by combining commercially available software with custom-written software to gain the specificity needed. Communications One of the primary benefits of chemical sensors is the ability to obtain chemical data in real time. Real-time data can be used to adjust and optimize experimental protocols and sampling design as data are obtained. Communication of these data is not now a problem for chemical sensors that are operated on board ship, or which are tethered to the ship by a hydrographic cable. Electromechanical cables with two or four conductors are almost universally available on research ships. These cables have a bandwidth great enough to transmit data at 4800 baud over 10,000 m. This should be sufficient for most ship-based chemical sensor applications. Far greater bandwidths can be obtained with fiber optic cables. These are not now in general use for oceanography, as there is limited demand for them, but submarine fiber optic cable and repeaters are available. Data communications is a much greater problem for remotely operated systems, such as those deployed on moorings in mid-ocean. A limited commercial capability for transmitting data via satellite is provided by Service Argos, Inc. The Argos system is flown aboard the NOAA Advanced TIROS-N satellites in a low earth orbit. As a result, Argos transmitters require little power, and complete units weigh as little as 165 grams. An Argos transmitter aboard an oceanographic buoy can send only 32 bytes of data to a satellite on each orbit, however. The mean number of satellite passes per day ranges from 7 at the equator to 28 at the North and South Poles. There is no capability for two-way communication, which would allow instructions to be transmitted to the sensor. Several other alternatives to the Argos system exist, but all have prob-
OCR for page 71
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies lems. The NASA GOES satellites, in geostationary orbits, now have the capability for continuous data transmission at 100 baud. This data rate will become 1200 baud in the near future. The transmitters require 30 watts of power, which is difficult to provide on a continuous basis for year-long time periods, but is feasible for intermittent use. This system can accommodate a limited number of users. Commercial communications satellites can also be used, but the transmitters again require large amounts of power and a large antenna. Communications will be a problem for retrieving data from remotely operated sensors as they become available. The benefits of near-real-time transmission of data from remote sites are clear. The data can be incorporated into oceanographic models to improve model predictions. Transmission of the data also eliminates the risk of losing the complete data set if the buoy is lost. For telemetered data, processing close to the sensor should be emphasized, to reduce the size of the data stream. For example, computed concentrations, rather than raw data, should be telemetered. Chemometric procedures programmed into the instruments can be used to help in this data reduction process. A new generation of satellites that provide the equivalent of a global cellular communications network will be necessary to take full advantage of chemical sensor systems.
OCR for page 72
Applications of Analytical Chemistry to Oceanic Carbon Cycle Studies This page in the original is blank.
Representative terms from entire chapter: