Environ. Sci. Technol. 2007, 41, 2296-2302
Wavelength Dependence of the Photochemical Reduction of Iron in Arctic Seawater LUIS M. LAGLERA AND CONSTANT M. G. VAN DEN BERG* Department of Earth and Ocean Sciences, University of Liverpool, Liverpool, L69 3GP, UK
low, natural background levels in seawater was determined at varying light intensities using natural, high-latitude sunlight, and filters were used to vary the spectral composition. Fe(II) concentrations were determined by flow analysis and detection by chemiluminescence (CL) in direct detection mode which enables detection within seconds. Redox Reactions of Iron in Natural Waters. The redox reactions of iron in water can be summarized as follows:
Fe(II) + An f Fe(III) + Bn
(1)
•+ • where A ) [O2]1, [O•2 + 2H ]2, [H2O2]3, [OH ]4, and B) [O2 ]1, [H2O2]2, [OH• + OH-]3, [O2]4
Chemiluminescence measurements of the photochemical reduction of iron in cold, high-latitude waters (79 °N) show that a significant fraction (20%) of the dissolved iron is reduced when exposed to sunlight. The reduction is immediately initiated and the transition to a steady-state concentration of ∼200 pM photochemical Fe(II) is achieved within ∼40 s. The photochemical Fe(II) is reoxidized to Fe(III) in less than a minute upon blocking the sunlight, much faster than expected, which is ascribed to reaction with photochemically produced oxidants. Using filters to block different ranges of the incident sunlight it was found that 35% of the photochemical Fe(II) was produced in the UV-B range (300-315 nm), 30% in the range 315360 nm, and 30% at higher wavelengths. Measurements of light attenuation as a function of depth indicate that photochemical Fe(II) at a depth of 5 m in high-latitude waters should amount to ∼10% of that at the surface. The fast kinetics modulate the paramount importance that photochemical reactions may have on the bioavailability of iron in surface waters.
Introduction Iron is an essential micronutrient that limits productivity in large parts of the oceans (1, 2). High-latitude, Southern Ocean waters are of particular importance because they constitute the largest ocean basin with iron-limited, low-productivity waters (3). Iron solubility and bioavailability are strongly controlled by its chemical speciation. Fe(III) is the thermodynamically most stable redox state of iron in pH 8 seawater with a solubility estimated at 10 pM due to the formation of insoluble hydroxide species (4). However, its overall solubility is increased by complexation with organic ligands (5) that dominate its equilibrium speciation in seawater (6-8). The chemical speciation of iron in seawater is complicated by its redox speciation. Fe(III) in seawater is reduced to Fe(II) both photochemically (9, 10) and chemically by reactive reducing substances such as the free superoxide radical (O•2 ) (11, 12) which itself is also produced by photochemical reactions involving dissolved organic matter (DOM). Fe(II) is oxidized in seawater by the oxygen/superoxide couple (1315), hydrogen peroxide (16), and the hydroxyl radical (17). The photochemical production of Fe(II) in seawater was demonstrated previously by illuminating seawater and added iron with a lamp in the laboratory (10, 11, 18). Here we determine for the first time the photochemical reduction of iron in sunlit arctic waters. The redox speciation of iron at * Corresponding author pnone: +44 151 794 4096; fax: +44 151 794 4099; e-mail:
[email protected]. 2296
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
Fe(III) + Cn f Fe(II) + Dn
(2)
where C ) (hv)1, [O•2 ]2, D ) [O2]2. The most reactive species is superoxide which is produced photochemically and is thought to control the redox speciation of iron in sunlit waters (11,19,20). The superoxide activity is affected by reaction with copper and DOM, both of which can therefore influence the redox speciation of iron. A concentration as low as 0.5 nM Cu could deplete the steadystate superoxide concentration by 2 orders of magnitude (11). Estimation of the steady-state ratio of Fe(II)/Fe(III) is complicated by the presence of organic complexing ligands because they affect the reaction rates and change the redox potential of the complexed species (21):
FeIIL + An f FeIIIL + Bn
(3)
FeIIIL + Cn f FeIIL + Dn
(4)
In addition to the superoxide reaction organic Fe(III) is reduced by direct photochemical reaction in which iron is electron acceptor by ligand to metal charge transfer (LMCT) (19). Organically complexed Fe(II) is susceptible to oxidation as a result of reaction with oxygen and any of the highly reactive radicals (19, 22). For instance, the superoxide radical is produced by photochemical reaction of natural chromophores (23):
L + hv f L*
(5)
L* + O2 f O•2 + L
(6)
Therefore, DOM plays a key role in the redox reactions and its effect depends on its composition.
Materials and Methods Reagents. The water used to prepare solutions (MQ water) was deionized by a Milli-Ro reverse osmosis and a Milli-Q deionizer system from Millipore. Hydrogen chloride, used for sample acidification and bottle cleaning, was purified by low-temperature distillation on a quartz condenser. An Fe(II) stock solution containing 4 mM Fe(II) was prepared from ferrous ammonium sulfate (AnalaR, BDH) in 0.2 M HCl. Fresh Fe(II) standards (10-5 M, pH 2.5) were prepared daily. Luminol (Sigma) (0.5 mM) was dissolved in 1 M ammonia buffer at a pH of 10.5. This solution was aged a week prior to beginning the experiments. Total dissolved iron and iron complexing ligands were determined by CSV with ligand competition against 2,3dihydroxynaphthalene (DHN) in the presence of 0.01 M 10.1021/es061994h CCC: $37.00
2007 American Chemical Society Published on Web 03/07/2007
Piperazine-N,N′-bis(2-hydroxypropanesulfonic acid) (POPSO) pH buffer (Fluka), set to pH 7.9, and 0.04 M bromate (Sigma), which had been purified as before (24). The DHN was added from a methanolic solution containing 0.013 M DHN to a final concentration of 20 µM (for dissolved iron) or 0.5 µM DHN (for iron speciation). The aliquots used for total dissolved iron analysis were kept at pH 2.1 and neutralized with ammonia (bidistilled) just before analysis. Copper and its speciation were determined by CSV with ligand competition against salicylaldoxime (SA) (25). Equipment and Analytical Procedures. Voltammetric instrumentation was a µAutolabIII from Ecochemie (Netherlands) with a VA693 mercury electrode stand from Metrohm (Switzerland). Chemiluminescence apparatus was FeLume (Waterville Analytical) using a multichannel peristaltic pump (model RP-1, Rainin). The concentration of Fe(II) in seawater was determined by continuous flow analysis with chemiluminescence detection (26). Sample and luminol were mixed continuously in a flow cell facing the photon counter. Instrument settings were modified as before (17) to enable continuous detection of Fe(II) in direct detection mode. The flow rate was 3 mL min-1 for the sample and 0.5 mL min-1 for the luminol, and the integration time was 250 ms. Connecting tubing was black Tygon and Teflon, wrapped with black plastic sheets to eliminate photochemical reactions during the transport phase. The length of the tube linking the Teflon bag and the detector was 2 m and the residence time of the water was 10 s. The chemiluminescence was monitored continuously on a computer. The limit of detection was determined from three times the value of the standard deviation of the signal and was 40 pM Fe(II). At the pH of natural seawater, all Fe(II) was expected to be labile and to include possible Fe(II) organic complexes (27, 28). The solar irradiance at the Ny-Alesund polar station was monitored with a Total Ultraviolet Radiometer (Eppley, TUVR) by the Alfred Wegener Institute (www.awi-bremerhaven.de) and integrated (1 min) in the range of 300 - 370 nm. The solar radiance was 20-25 W m-2 during the experimental period. Sunlight spectra were registered with a USB2000 spectrophotometer (Ocean Optics) with a 15 m optical fiber, which was used on land and in the water to a depth of 5 m. A cosine-corrected irradiance probe (CC-3-UV, Ocean Optics) was fitted at the end of the fiber to collect radiation over a range of ∼180° and eliminate problems associated with the geometry of the light collection during sampling. The system was calibrated in the field over a wide range of solar intensities against a PUV-500 radiometer. Radiometer outputs represent energy integrated over a 1 nm width range at 305, 320, 340, and 380 nm. The calibration plot of these values versus equivalent integrations of the spectrophotometer readings was linear in all cases tested. Seawater Irradiation Experiments. Water (salinity 34; pHNBS 8.10 at 20 °C corrected to an in-situ pHF 8.35 at 5 °C (29, 30)) was collected by peristaltic pump in Kings Bay fjord close to the research station of Ny-Alesund (79°N, 11°6′E) on the island of Svalbard in June 2002. The water was filtered on-line (0.2 µm filtration cartridge) and stored overnight in a dark 5 L carboy at ambient temperature (∼4 °C) to deplete any Fe(II) and other transient species present in the water. One hour before the start of the experiment the water was transferred to a 500 mL, FEP-Teflon bag (VueLife, American Fluoroseal Corporation). Checks showed the bags to be transparent for the entire spectrum tested (300-800). The Teflon bag (surface area 180 cm2) was placed flat on a panel immediately outside an open window of the laboratory and exposed to solar radiation, with and without filters to selectively remove certain wavelengths from the incoming solar radiation. Water was pumped continuously from the
bag. The water temperature during the experiment was approximately equal to the air temperature (around 3-4 °C). The laboratory was unheated and the windows remained open to avoid temperature changes. Plastic filters were placed in front of the Teflon bag to block light of wavelengths shorter than the filter cutoff. Two of the filters were acetates (labeled as A and B), and a third one was a Mylar polyester film; a black sheet of polyethelene was used for complete light blocking. The light cutoff wavelength of these filters was calibrated using the spectrophotometer. Mylar (cutoff at 315 nm) is often used to remove the UVB component of light. At higher wavelengths, the transmittance was >85%. Acetate filters labeled as A and B had cutoff wavelengths of 354 and 362 nm. The transmittance at higher wavelengths was 90% for filter A and ∼100% for B. Calibration of the Chemiluminescence Response for Fe(II). The CL response for Fe(II) was calibrated by 1 nM Fe(II) additions to seawater in a shielded polyethylene bottle at ambient temperature (5 °C) while pumping through a short length of black tubing to the detector where it was mixed with luminol immediately prior to the measurement. Correction was made for the decay of Fe(II) during the transport (10 s) to the detector (at rates in the range 2.5-3 × 10-3 s-1). The procedure was repeated immediately after each set of analyses.
Results Preliminary Experiments. The concentration of dissolved iron in the fjord water collected at Ny-Alesund was found to be 0.9 ( 0.1 nM. The concentration of Fe(III)-binding ligands was 1.6 ( 0.1 nM with a complex stability of log K′Fe′(III)L ) 11.8 ( 0.4. The ratio of organic over inorganic iron (RFeL ) [FeL]/[Fe′]) in this water was therefore ∼103, indicating that the iron(III) speciation in equilibrium condition was dominated by organic complexation, as 99.9% of the iron was organically complexed.The dissolved copper concentration in the same water was 1.3 nM and the copper complexing ligand concentration 16 nM (log K′CuL ) 12.8); the ratio of organic over inorganic copper (RCuL) was therefore ∼104. The concentrations of copper and iron are low for coastal waters but in line with expectations for deeply mixed northern Atlantic Ocean water (31). The CL system was flushed with aged fjord water until a stable reading of ∼500 counts was obtained, mainly due to leakage of ambient light. When the luminol was pumped through, the signal increased to a stable value of ∼850 counts. The difference is equivalent to 0.08 nM Fe(II) which could be from residual Fe(II), possibly stabilized by organic complexation, or from light emission from the interaction of luminol with fjord water components other than Fe(II) (27, 28). In spite of a lack of direct evidence, it is thought that Fe(II) organic complexation exists in seawater (32, 33) based on the apparent stability of Fe(II) during its collection from deep waters and slower than predicted Fe(II) oxidation rates. The CL response is known to include Fe(II) bound with organic ligands (27, 28). A return to the same, stable (dark), background response after oxidation of photochemical Fe(II) during the experiments indicated that bleaching effects were insignificant, suggesting that the background response was not due to stabilized Fe(II) complexes. Various compounds are known to give a CL response. The experimental conditions (pH, temperature) were therefore kept constant to eliminate variations in the background. Cobalt (Co) is thought to give a CL response in conditions similar to those used here for iron (34). However, nanomolar additions of Co, much greater than natural background levels (35), did not give a CL response, presumably because the analytical conditions were not optimal for cobalt, which therefore did not interfere with the iron detection. The VOL. 41, NO. 7, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2297
FIGURE 1. Photochemical production of Fe(II) in arctic seawater by sunlight. (A) Fe(II) production when seawater is exposed to, and shielded from, sunlight at various wavelengths. Raw data from the CL apparatus are shown. Vertical arrows show the time when a filter was placed or removed. The three examples were not run consecutively so time is just for the purpose of comparison. (B) Energy effect: relationship between the photochemical Fe(II) concentration and the energy in the UV range of the sunlight. hypothetical contribution of other species was not further investigated. Throughout this work the concentration of photochemical Fe(II) was measured above the stable, 850 count, background. Irradiation Experiments. The experiment was started by pumping the water from the Teflon bag (covered with black plastic to block the sunlight) for several minutes until a stable signal (850 counts) was obtained, after which the cover was removed. The response of the CL detector increased immediately, reaching a maximum of approximately 1700 counts within ∼40 s in response to the photochemical generation of Fe(II) (Figure 1A). The maximum was followed by a decrease during ∼1.5 min until the response stabilized at ∼1500 counts and then remained constant for at least 10 min. A return to the original background count when the light was removed showed that the photochemical Fe(II) was immediately oxidized. The steady-state concentration of photochemical Fe(II) is therefore the balance of Fe oxidation and reduction. The oxidation was completed within 50-60 s. The steady-state concentration of the photochemical Fe(II) was 180 pM, equivalent to 20% of the dissolved iron concentration. Effect of the Light Intensity. Photochemical Fe(II) production was measured at various times of the day over several days in an attempt to relate the photochemical generation of Fe(II) to the light intensity. The steady-state concentration of photochemical Fe(II) was found to be linearly related to the radiation energy in the wavelengths of 300-370 nm (Figure 1 B). Extrapolation to low light intensities gives an intercept with the X-axis close to the XY-axis intercept, indicating that very little activation energy is required to initiate the iron reduction: i.e., iron(II) is produced already at a very low light intensity. A maximum was not achieved in these experiments indicating that the steady-state fraction of iron photochemically converted to Fe(II) continues to increase at greater irradiance, as occurring at lower latitudes. 2298
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 7, 2007
FIGURE 2. Effects of varying the wavelength range of the sunlight on the Fe(II) production. (A) Fe(II) as percentage of that produced by full sunlight. (B) Fe(II) production rates for different wavelength ranges. Numbers between parentheses correspond to the number of experiments. Bars are standard deviations. Ranges between parentheses correspond to the solar wavelengths added to the incoming radiation by the removal of the filter. The insert shows the linearity of ln [Fe(II)] as a function of time, which was used to calculate rate constants. Effect of Varying the Wavelength Range of the Light on the Photoreduction of Iron. The wavelength range of the incident light was varied by using filters to block the radiation at wavelengths shorter than the cutoff wavelength. Water was alternately exposed to filtered and unfiltered light and the concentration of Fe(II) was monitored. The steady-state Fe(II) concentration was found to decrease when light was blocked at a higher wavelength in the UV range, i.e., when a wider range of UV light was filtered out (Figure 2A). Examples of the CL variation are shown in Figure 1A, and the % Fe(II) suppression compared to sunlight is shown in Figure 2A. Steady-state CL signals were equivalent to 120 pM Fe(II) for the Mylar filter (cutoff at 315 nm) and 70 pM Fe(II) for filters A and B (cutoff at around 360 nm). The transition time to lower the Fe(II) signal was ∼1 min for filters A and B and ∼2 min for Mylar. The photoreduction of iron was lowered by 30% by the Mylar filter, which means that UV-B light of wavelengths
