Automation of flow injection gas diffusion-ion chromatography for the nanomolar determination of methylamines and ammonia in seawater and atmospheric samples

The automation and improved design and performance of Flow Injection Gas Diffusion-Ion Chromatography (FIGD-IC), a novel technique for the simultaneous analysis of trace ammonia (NH3) and methylamines (MAs) in aqueous media, is presented. Automated Flow Injection Gas Diffusion (FIGD) promotes the selective transmembrane diffusion of MAs and NH3 from aqueous sample under strongly alkaline (pH > 12, NaOH), chelated (EDTA) conditions into a recycled acidic acceptor stream. The acceptor is then injected onto an ion chromatograph where NH3 and the MAs are fully resolved as their cations and detected conductimetrically. A versatile PC interfaced control unit and data capture unit (DCU) are employed in series to direct the selonoid valve switching sequence, IC operation and collection of data. Automation, together with other modifications improved both linearily (R2 > 0.99 MAs 0-100 nM, NH3 0-1000 nM) and precision (<8%) of FIGD-IC at nanomolar concentrations, compared with the manual procedure. The system was successfully applied to the determination of MAs and NH3 in seawater and in trapped particulate and gaseous atmospheric samples during an oceanographic research cruise.


Introduction
Nitrogen, a bio-essential element in the marine environment [1,2], is found in a variety of inorganic and organic forms in oxic seawater, ranging from the thermodynamically most stable species, nitrate (NOr), to the reduced compounds such as ammonia (NH3) and its mono-, di-, and tri-methylamine derivatives (CH3NH. (CH3)2NH and (CH3)3N abbreviated MMA, DMA and TMA respectively). Methylamines (MAs), like NH 3 are polar, volatile, water soluble species which undergo extensive hydrogen bonding to form basic solutions (pK 9.25-10.77). Due to their low molecular weight, ability to participate in phase transfer processes [3][4][5][6] and importance in marine nitrogen fertility [1,7], detoxification and osmoregulation [8][9][10], NH 3 and MAs are widely distributed and dynamic within the marine environment.
Due to their volatility, NH 3 and the MAs (boiling points -33"4-7"4C) are also capable of gaseous evasion across the air-sea, interface, thus introducing alkali and reduced nitrogen into the troposphere [3,4,11]. Here, through processes such as their dissolution into cloud or rainwater, reaction with H2SO4 aerosol particles and photochemical oxidation to NOx, NH3 and the MAs may influence the redistribution of both nitrogen and sulphur, the acid-base chemistry of the atmosphere and, ultimately, climatic parameters such as the number density and chemical composition of cloud condensation nuclei [3,4,[11][12][13].
Methylamines are of additional importance due to the conversion of secondary amines (for example DMA) to their N-chloro-derivatives in chlorine disinfected wastewaters [ 14-1 and the implication of secondary and tertiary amines in the synthesis of carcinogenic nitrosamines in aqueous media, air, soils and foodstuffs [15]. Aspects of the marine biogeochemistry of NH 3 and MAs have been studied over many years (for example [3,5,8]). However, understanding their distribution and transformations has been restricted by the absence of an analytical technique capable of their selective quantification at the nanomolar levels typical of the marine environment.
Ammonia is often determined in natural waters potentiometrically by ion-selective electrodes [16,17], or colorimetrically, typically by a version of the indophenol blue method [18]. However, away from bacterially active, anoxic or anthropogenically influenced regions, concentrations of NH 3 often fall below the sensitivity of these techniques (,-0"1 gM). Only recently have cathodic stripping voltammetry [19] and 0-phthalaldehyde (OPA) [20] fluorescence techniques with limits of detection < 10 nM been reliably applied to the analysis of ammonia in pristine and open oceanic waters.
Methylamines, meanwhile, are normally studied by Gas or High Performance Liquid Chromatography (GC or HPLC). In practice, many GC techniques suffer from peak asymmetry [21][22][23], ghosting phenomena [24][25][26] and detector response quenching [21,23,27], while HPLC techniques require derivatization of MAs before detection. While derivatization is advantageous in overcoming the often problematic polar, volatile nature of the MAs, there is no single derivatizing agent available for primary, second and tertiary amines. Only in ion chromatography (IC) are these problems overcome and it is possible to simultaneously analyse ammonia and MAs (as solvated ammonium (NH2) and methylammonium cations) [21,28,29]. S. W. Gibb el al. Automation of flow injcction gas diffusion-ion chromatography Injection Gas Diffusion coupled to Ion Chromatography (FIDG-IC), which exploits the advantages of IC and permits the simultaneous analysis of NH 3 and the MAs at nanomolar levels by a single analytical technique in natural waters [30]. This paper reports on the improvement, automation and computer interfacing of the technique which has increased its precision and reliability. The applicability of the technique is demonstrated through the analysis of MAs and ammonia in seawater and atmospheric samples during a research cruise in the northwestern Indian Ocean.

Reagenls
Ammonia and MA stock solutions (0" 10 M) were prepared from their hydrochloride salts (Fluka). Single and mixed standards were prepared daily from these stocks. Internal standards (ISs)cyclo-propylamine (c-PA) and secbutylamine (s-BA) were prepared from serial dilution of laboratory grade reagents (Sigma, UK). Cyclo-PA was chosen as an IS since its occurrence in the natural environment has not been reported. The alkaline-chelating reagent, 1"1 M ethylene diamine 5V,V,V',V', tetra-acetic acid (EDTA)/0" 11 M NaOH was prepared t?om the tetra-sodium salt of EDTA and ACS grade sodium hydroxide pellets (Sigma, UK). Eluent (40 mM MSA) was routinely prepared via a 1"0 M stock t?om 'Aristar' grade concentrate (BDH, UK).
Water taken fi'eshly from a Milli-Q Water Purification System (Millipore, UK) with a specific resistivity of > 18 Mf and further passed through a sealed ionexchange column packed with Amberlite IR-120-t-(proton tbrm) was used as the diluent or solvent in the preparation of standards and reagents. Further ionexchange was used to reduce the intertrence of alkali metals (Na + and K +) and the contribution of blanks (NH3 and MAs).

Ion chromalograph and suppression syslem
Chromatography was pertbrmed on a Dionex DX-100 I C The diffusion and stripper cells were custom designed and fabricated from milled perspex [30]. Each composite module consists of a pair of mirror image blocks into which a zig-zag shallow rectangular cross-section channel was cut (length 1004 mm, width 2"0 mm, depth ,-0"1 mm).
The two blocks were secured using stainless-steel screws and reproducibly and uniformly tightened using a calibrated torque-limiting screwdriver.
Microporous PTFE was used as a the gas-exchange membrane (Goretex MF/002/PM-pore size ItM, thickness 0"076 mm, porosity 78; W. L. Gore, UK). Supplied in sheet form, this material was cut to a template while sandwiched between sheets of paper to facilitate easy handling. A full analytical description and evaluation of the original FIGD-IC technique has previously been published [30]. A series of normally closed solenoid operated valves were used to control and direct liquid flow in the FIGD system (Biochem Valve Corp., USA).

Software, PC and automation
The analogue output signal (10 ItS range; 10% oflet/l V output) from the IC was converted to digital (2 Hz) using a Philips PU6031/10 data collection unit (DCU) and then processed using ATI  The control input is connected to a Schmitt NAND gate (fbur of which are available in one CMOS 4093 series package). The output of the gate is connected to a high current VMOS field effect transistor which acts as the switch fbr turning the solenoid on and off. When the input reed relay is open, the two inputs of the NAND gate are at logical due to the two resistors which act as 'pull-ups' to the 5 V line. The output of the NAND gate will therefbre be at logical 0, the transistor will be turned off, and the solenoid will not be energized. The diode in parallel with the solenoid prevents the build-up of back e.m.f, when the solenoid is turned off.
When the reed relay contacts close, after a software command, one input of the NAND gate will go to logical 0 thus fbrcing the output to go to logical 1. This turns on the transistor and the solenoid will be energized. The action is the same if the 'contact closure' is performed by an open collector transistor, or the control line is fbrced to logical 0 by a TTL logic level. The interface is therefore quite versatile. The use of a Schmitt NAND gate ensures positive on-off switching, even if the control input changes slowly. The built-in hysteresis of this gate prevents solenoid 'chatter' during switching, which would generate considerable electrical and mechanical noise.
The development of the circuit for use with values V2 and V3 is shown in figure 3, where pairs of solenoids are operated simultaneously in a combination of two, twoport valves. These are interconnected to control the enrich or flush/discharge cycle, and enable the flow to be switched between cycles by a single timed event in the software.
When the control contacts are open, solenoids S1 and $2 are released but, due to the inverting action of a second VMOS field effect transistor, solenoids $3 and $4 are energized. It follows that when the control contacts close, S1 and $2 are energized and $3 and $4 are released.
Remote IC injeclion: To facilitate the automatic injection of the sample, another simple interface is required. This interface must allow the control signal to operate a pneumatic valve on the I C which is triggered by linking two electrical contacts. While this could be performed by the control contacts directly, it was tilt useful to preserve the 'universal' nature of the interfiace and arrange for the link to be made by a VMOS field effect transistor. This could then be controlled by an open collector or TTL logic level as befbre. The circuit of a suitable interface is shown in figure 3 (note that this is very similar to figure 2).   " normalized wrt c-PA (IS); n/a not applicable.
graphically resolved using isocratic elution (40mM MSA). The monovalent cations are conductimetrically detected with chemical suppression of" the background signal. The analogue signal (which is proportional to the analyte concentration) is digitized by the DCU and collected and processed by the 4880 software system.
Since FIGD-IC response has previously been shown to relate linearly to difl'usion time (Re> 0"99, 1-60 min [2 l-l), difl'usion times can be selected in accordance with analytes concentration. As a consequence, with the appropriate switching sequence, utilization ofditi'usion or enrichment times in excess of the 15-min chromatographic run time, permits enrichment of one sample and chromatography of the preceding sample to proceed in parallel [5]. This significantly increases sample throughput and operational efficiency.
Chemical suppression systems, such as the CSRS employed here, greatly enhance signal to noise ratio by effectively decreasing background eluent conductivity, increasing relative analyte detector response and eliminating system peaks by removing counter ions. The result is a significant improvement in analyte detection limits. CSRS is a high capacity, automatic suppression system in which the use of detector cell effluent as a water source eliminates the need tbr chemical regencrant solutions or ion-exchange cartridges. With a dead volume of only 30 t.tl analyte dispersion eftcts are minimal.
Since all reagent and sample bottle headspace is fed with acid-scrubbed air tied through PTFE lines, atmospheric contamination is minimized and reagent litie is prolonged.
Using a second diffusion or 'stripper' cell (SC in figure   1) minimizes the NH and MAs' content of the EDTA/ NaOH reagent betbre its addition to the sample. Eluent and FIG1) acceptor solution (both 40mM MSA) are prepared concurrently, ensuring the two are matched and thereti)re minimizing I C base-line perturbation upon acceptor injection. l'he two internal standards incorporated into the analyti-cal scheme were used to monitor FIGD-IC perfbrmance (figure 1). Sec-BA, added to the acceptor (typically 10 laM), to monitor IC stability and reproducibility, while c-PA spiked into the sample was generally used to quantif) determinant concentrations through predetermined relative response factors. It was also possible to determine a range of other low molecular weight alkylamines from aqueous samples [21].

Calibration, precision and sensitivity
The automated FIGD-IC system gave a highly linear response to NH and MAs in both standards and spiked seawater with relatively good precision in the concentration range of" interest (see table 1). Both R 2 and RSD coefficients show, in general, an improvement over those achieved with the manual system. This is particularly important in the trace determination of MAs.
On-column detection limits fbr IC alone were 0"20 ng NH-, 0"37 ng MMA, 0"54 ng DMA, 0"95 nf TMA and 0"72 ng c-PA for a 200 l-tl injection volume [21]. Limits of detection fbr the coupled FIGD-IC system are dependent upon the diffusion time employed, base-line stability, the relative abundance of analytes and the influence of the system blank. Since NH, in particular, gives a significantly non-zero blank, this makes trace analysis of NH more difficult than tbr the MAs. In practise, improvements in MA detection limits (but not NH3) are achieved using the automated system with a 40min difl'usion time (table 1). Detection limits are improved most noticeably tbr DMA and TMA, since these occur in regions of greater base-line stability.

Applications
The automated FIGI)-IC system was deployed fbr the analysis of NH and MAs during Cruises 210 and 212 of  [31] and diluted in MQ.water (to 50 ml) and analysed by FIGD-IC using a 20min diffusion time. Both particulate and gaseous concentrations were background corrected using 'blank' filters treated in the same manner as the sampled filters but without air pumping (see figure 6).

Results
The automated FIGD-IC system was successfully applied to the determination of NH 3 (table 2). However, in atmospheric samples, the relative abundance of NH 3 with respect to MMA made quantification of MMA difficult since it appeared only as a shoulder to the NH 2 peak (figure 6). Furthermore it was not possible to quantify DMA in the gaseous phase due to an intertiring peak arising fiom processing the acid soaked paper filter. In general, NH 3 was the dominant species occurring at 5-1000 times greater concentrations than those of any MA (whose levels occasionally fall below the detection limits of the system). Monomethylamine was generally observed to be the dominant MA in Indian Ocean seawater, while TMA was normally detected at concentrations of <5 nM.  generally observed in the regions of the thermocline and greatest primary production (shown as 'Temperature' and 'Chlorophyll a' concentration respectively in figure 7).

Comparison with other analytical techniques
Techniques tbr the analysis of nanomolar concentrations ot" MMA, DMA and TMA generally require prolonged preconcentration times (up to 36 h) [24,32,33], also they are labour intensive [21,32,33] or they are. unsuitable tbr use on ship [32]. FIGD-IC, on the other hand, allows determination of up to four samples per hour and is also less susceptible to contamination than the other techniques. The prerequisite ot'a chromatographic step tbr simultaneous determination ot" NHa and MAs, results in a considerably lower sample throughput than may be achieved tbr a single determinant. Recently it has been possible to determine NH by OPA-fluorimetry using flow injection principles and a gas diffusion cell [20]. This

Conclusion
The automated FIGD-IC system proposed in this paper permits the near real-time determination of MAs and NH 3 in seawater and aqueous extracts of atmospheric gaseous and particulate samples. Simultaneous analysis of nanomolar concentrations are possible without the need fbr derivatization schemes or lengthy preconcentration procedures. While the manual FIGD-IC system required constant supervision (lbr switching valves, injecting samples etc.), automation significantly reduced this need.
Automation improved the analytical reproduciblity and sample throughput of the system by increasing precision and accuracy valve switching and by reducing the scope [br human error.
The flexible nature of the automation control system described allows considerable scope for broader applications of the FIGD-IC and also for other automated analytical applications, for example other natural waters such as rainwater and fresh waters.