Continuous High-Frequency Dissolved O 2 /Ar Measurements by Equilibrator Inlet Mass Spectrometry

Continuous High-Frequency Dissolved O2/ArMeasurements by Equilibrator Inlet MassSpectrometryNicolas Cassar,*,† Bruce A. Barnett,† Michael L. Bender,† Jan Kaiser,‡ Roberta C. Hamme,§ andBronte Tilbrook|

Department of Geosciences, Princeton University, Princeton, New Jersey 08544, School of Environmental Sciences,University of East Anglia, Norwich, NR4 7TJ, U.K., School of Earth and Ocean Sciences, University of Victoria, P.O.Box 3055 STN CSC, Victoria, British Columbia, V8W 3P6, Canada, and Commonwealth Scientific Industrial ResearchOrganisation (CSIRO) Wealth from Oceans Research Program and Antarctic Climate and Ecosystem CooperativeResearch Center, Hobart, 7001 Tasmania, Australia

The oxygen (O2) concentration in the surface ocean isinfluenced by biological and physical processes. Withconcurrent measurements of argon (Ar), which hassimilar solubility properties as oxygen, we can removethe physical contribution to O2 supersaturation anddetermine the biological oxygen supersaturation. Biologi-cal O2 supersaturation in the surface ocean reflects thenet metabolic balance between photosynthesis and res-piration, i.e., the net community productivity (NCP). Wepresent a new method for continuous shipboard mea-surements of O2/Ar by equilibrator inlet mass spectrom-etry (EIMS). From these measurements and an appro-priate gas exchange parametrization, NCP can be esti-mated at high spatial and temporal resolution. In the EIMSconfiguration, seawater from the ship’s continuous intakeflows through a cartridge enclosing a gas-permeable mi-croporous membrane contactor. Gases in the headspaceof the cartridge equilibrate with dissolved gases in theflowing seawater. A fused-silica capillary continuouslysamples headspace gases, and the O2/Ar ratio is measuredby mass spectrometry. The ion current measurements onthe mass spectrometer reflect the partial pressures ofdissolved gases in the water flowing through the equilibra-tor. Calibration of the O2/Ar ion current ratio (32/40) isperformed automatically every 2 h by sampling ambient airthrough a second capillary. A conceptual model demon-strates that the ratio of gases reaching the mass spectrom-eter is dependent on several parameters, such as thedifferences in molecular diffusivities and solubilities of thegases. Laboratory experiments and field observations per-formed by EIMS are discussed. We also present prelimi-nary evidence that other gas measurements, such as N2/Ar and pCO2 measurements, may potentially be performedwith EIMS. Finally, we compare the characteristics of theEIMS with the previously described membrane inlet massspectrometry (MIMS) approach.

The departures from saturation in dissolved oxygen (O2)concentration in ocean surface waters are the consequence of

physical and biological processes. O2 and the inert gas argon(Ar) have similar solubility properties. Measurements of Ar cantherefore be used to remove the disequilibrium of dissolvedO2 in waters of the oceanic mixed layer associated with heatfluxes, bubble injections, and variations in atmosphericpressure.1-3 The remaining O2 disequilibrium, “the biologicalO2 supersaturation”, reflects the competing influences of mixed-layer net community production (NCP), vertical mixing, andgas exchange. NCP corresponds to the difference betweenrates of oxygenic photosynthesis and respiration by thebiological community. Biological oxygen is supersaturated(undersaturated) within the mixed layer if community photo-synthesis/respiration is greater (less) than 1 (correcting forpotential bias from entrainment, upwelling, or eddy diffusionof oxygen-undersaturated waters). With a gas exchangeparametrization,4-8 net biological oxygen flux at the air-seainterface, and net community productivity (NCP) within the mixedlayer, can be estimated. Because of the close stoichiometric linkbetween O2 and organic carbon fluxes,9 and the approximatelyweek-long residence time of oxygen within the mixed layer,NCP derived from O2/Ar measurements provides a uniquemeasure of upper ocean carbon fluxes over the week or soprior to measurements.

Several studies have reported discrete NCP estimates derivedfrom in situ O2/Ar measurements.1,3,10-13 Continuous measure-ments offer the opportunity to capture the heterogeneity in oceanic

* To whom correspondence should be addressed. E-mail: [email protected]. Phone: 609-258-7435. Fax: 609-258-1274.

† Princeton University.‡ University of East Anglia.§ University of Victoria.| Commonwealth Scientific and Industrial Research Organisation.

(1) Craig, H.; Hayward, T. Science 1987, 235, 199–202.(2) Emerson, S. J. Geophys. Res., [Oceans] 1987, 92, 6535–6544.(3) Spitzer, W. S.; Jenkins, W. J. J. Mar. Res. 1989, 47, 169–196.(4) Wanninkhof, R.; McGillis, W. R. Geophys. Res. Lett. 1999, 26, 1889–1892.(5) Wanninkhof, R. J. Geophys. Res., [Oceans] 1992, 97, 7373–7382.(6) Ho, D. T.; Law, C. S.; Smith, M. J.; Schlosser, P.; Harvey, M.; Hill, P. Geophys.

Res. Lett. 2006, 33, L16611.(7) Nightingale, P. D.; Liss, P. S.; Schlosser, P. Geophys. Res. Lett. 2000, 27,

2117–2120.(8) Nightingale, P. D.; Malin, G.; Law, C. S.; Watson, A. J.; Liss, P. S.; Liddicoat,

M. I.; Boutin, J.; Upstill-Goddard, R. C. Global Biogeochem. Cycles 2000,14, 373–387.

(9) Shulenberger, E.; Reid, J. L. Deep-Sea Res., Part A 1981, 28, 901–919.(10) Hendricks, M. B.; Bender, M. L.; Barnett, B. A. Deep-Sea Res., Part I 2004,

51, 1541–1561.(11) Reuer, M. K.; Barnett, B. A.; Bender, M. L.; Falkowski, P. G.; Hendricks,

M. B. Deep-Sea Res., Part I 2007, 54, 951–974.(12) Cassar, N.; Bender, M. L.; Barnett, B. A.; Fan, S.; Moxim, W. J.; Levy, H.;

Tilbrook, B. Science 2007, 317, 1067–1070.(13) Hamme, R. C.; Emerson, S. R. J. Mar. Res. 2006, 64, 73–95.

Anal. Chem. 2009, 81, 1855–1864

10.1021/ac802300u CCC: $40.75 2009 American Chemical Society 1855Analytical Chemistry, Vol. 81, No. 5, March 1, 2009Published on Web 02/04/2009

community production. We present a new method for measuringcontinuously the ratio of dissolved O2/Ar in seawater byequilibrator inlet mass spectrometry (EIMS). Seawater ispumped through an equilibrator cartridge, and the gases inthe headspace of the cartridge assume partial pressures whichreflect their dissolved concentrations. The headspace gasesthen flow through a capillary tube into the mass spectrometer,which measures the O2/Ar ion current ratio, among otherproperties. EIMS builds on the established methodology ofmembrane inlet mass spectrometry (MIMS).14-16 However,there are also important differences, which give each method itsown niche. As opposed to MIMS, EIMS does not require liquidstandards (e.g., air-equilibrated water) for calibration of ion currentratios. The measurements performed with a MIMS instrumentcan only be as accurate as the liquid standards.15,16 Calibrationof the EIMS instrument is achieved by periodically samplingambient air through a second capillary. EIMS also requires lessperipheral instrumentation. The response time of EIMS is,however, significantly longer than MIMS. EIMS has now beensuccessfully deployed on several cruises. Examples of the continu-ous in situ O2/Ar observations are reported below.

MATERIALS AND METHODSDescription of Equilibrator Inlet Mass Spectrometry. In

the shipboard configuration of the EIMS, seawater from the ship’sunderway system flows continuously through a gas equilibrator.Relatively rapid gas exchange and equilibration of the dissolvedgases with the headspace of the equilibrator is ensured by thelarge surface area of the small, gas-permeable, hollow tubespacked into the equilibrator (see description of equilibrator below).The seawater flows perpendicular to, and outside of, the hollowfibers. The equilibrated gas phase in the headspace of thecartridge is continuously sampled through a fused-silica capillary(2 m in length and 0.05 mm diameter) for monitoring of gas ratioson a quadrupole mass spectrometer. The ion current measure-ments made by the mass spectrometer depend on the partialpressure of dissolved gases in the water flowing through theequilibrator. Because of the potentially large changes in roomtemperature on a ship, the ion source and mass filter of thequadrupole are temperature-controlled at 50 °C (±0.2 °C) with acustom-built, temperature-regulated, enclosure.

The quadrupole mass spectrometer used in this study is aPfeiffer Prisma model QMS 200 M1 with a Prisma gas-tight ionsource with yttrium oxide coated iridium filament. The sourceemission is 0.50 mA. The detector is a continuous dynodesecondary electron multiplier run at 1400 V amplification. Themass resolution is 50 (QuadStar software setting) with a dwelltime of 50 ms. The pumping speed of the turbo pump is 60 L s-1.Pressure is measured with a compact full-range Pirani/coldcathode gauge (PKR 251).

Below is a more detailed description of the various componentsof the equilibrator inlet.

Upstream of the Equilibrator. A multiple-step filtrationsystem upstream of the equilibrator ensures that the equilibrator

does not clog. First, coarse filtration is performed with aninline reusable hose filter (pore size 500 µm) in the high flowrate seawater line (3-5 L min-1) (Figure 1). The high flow rateminimizes the rise in seawater temperature and decreases theresidence time of the seawater in container B (Figure 1). A fractionof the seawater is pumped through the equilibrator at a nominalrate of 100 mL min-1 with a gear pump. Fine filtration isperformed upstream of the equilibrator with a custom-madefilter sleeve with 100 and 5 µm pore size polypropylene felt onthe outside and inside, respectively (100 µm/5 µm filter bag,1.5 in. width × 12 in. length, McMaster-Carr cat. no. 98315K99).Tygon silver antimicrobial tubing is used to inhibit biologicalgrowth within the lines upstream of the equilibrator.

Equilibrator Description. The equilibrator is a small (26.9mm × 73.1 mm), commercially available cartridge (http://www.liquicel.com, MicroModule 0.75 × 1), similar to the one usedby Hales et al.17 for pCO2 measurements by nondispersiveinfrared absorbance detection. Discs of hollow fiber arrays ofgas-permeable and hydrophobic material (polypropylene/epoxy) are tightly packed in a polycarbonate housing. This

(14) Kaiser, J.; Reuer, M. K.; Barnett, B.; Bender, M. L. Geophys. Res. Lett. 2005,32, L19605.

(15) Kana, T. M.; Darkangelo, C.; Hunt, M. D.; Oldham, J. B.; Bennett, G. E.;Cornwell, J. C. Anal. Chem. 1994, 66, 4166–4170.

(16) Tortell, P. D. Limnol. Oceanogr., Methods 2005, 3, 24–37.(17) Hales, B.; Chipman, D.; Takahashi, T. Limnol. Oceanogr., Methods 2004,

2, 356–364.

Figure 1. The large seawater reservoir (A) sits in a sink. After goingthrough an inline coarse filter (500 µm pore size), seawater flowsinto the inner reservoir (B) at a rate of 3-5 L min-1 (large arrow).Most of the water running into B overflows into A, which is used asa water bath thermostatted to the temperature of ambient seawater.A small fraction (100 mL min-1) of the high flow rate is pulled with agear pump through a filter sleeve (C), with 100 and 5 µm pore sizeon the outside and inside, respectively. From the gear pump, theseawater flows through the equilibrator (D). The equilibrator sits inreservoir A to keep its temperature identical to that of the incomingseawater. A capillary, attached to the headspace of the equilibrator,leads to a multiport Valco valve. This valve alternates betweenadmitting gas from the equilibrator and ambient air to the quadrupolemass spectrometer. An optode (not shown) in container B measurestotal oxygen saturation. Also not shown is a water flow meter locateddownstream of the equilibrator and thermocouples monitoring tem-peratures throughout the system.

1856 Analytical Chemistry, Vol. 81, No. 5, March 1, 2009

configuration provides a large surface area (392 cm2) for rapidgas exchange with a small headspace (2 mL), leading torelatively rapid gas equilibration between the headspace of thecartridge and the circulating seawater. The fibers have aninternal diameter of approximately 200 µm and an outerdiameter of approximately 300 µm, with a nominal wallthickness of 50 µm. Twenty-five percent of the surface ispermeated with pores with an effective size of 0.04 µm. Gasesflow to the mass spectrometer through a capillary inserted intothe headspace of the equilibrator.

Air Calibration Configuration. Relative to dissolved O2/Arin surface seawater, the atmospheric O2/Ar is essentiallyconstant. The ion current ratio (32/40) recorded when the massspectrometer is sampling the equilibrator can be calibrated tocompute the concentration ratio, by admitting ambient airthrough a second capillary. The capillary going to the massspectrometer is connected to a Valco valve (HPLC Streamselector valve with 1000 psi rating, cat. no. C5-1306EMH2Y)that switches between two capillaries of identical dimensions,one coupled to the equilibrator, and one sampling ambient air.The air calibrations are implemented by switching the valvebetween the two capillaries. Switching occurs automaticallyevery 2-4 h with a computer-controlled actuator, and ambientair is analyzed for 10 min.

Downstream of Equilibrator and Ancillary Measurements.The water flow rate is monitored continuously with a Ryton50-500 mL min-1 flow meter (Cole Parmer cat. no. C-32703-52) downstream of the equilibrator. A continuous dissolved O2

monitor (Aanderaa optode model 3835), calibrated with Winklertitrations,18,19 sits in container B (Figure 1), next to the fine filtersleeve, and measures total oxygen concentration.

Software Support. A single computer simultaneously logsdata from several instruments. The ion currents measured by the

mass spectrometer (Pfeiffer Prisma quadrupole) are recorded withthe Quadstar 32-bit software with process control module andquantitative analysis module. The optode’s oxygen saturation andconcentration observations are recorded with Oxyview (Aanderaasoftware supplied with the optode). Laboratory temperature,measured with several thermocouples, and the seawater flow rateare recorded with National Instruments LabVIEW. The latterprogram also controls the valve switching from water to ambientair for calibrations.

A MATLAB script is used to consolidate all the parametersmeasured by these various programs, to average them into 2 minintervals, and to plot them along with ancillary data, in near realtime, in a figure with multiple panels (Figure 2). The mainadvantage of this approach is that the EIMS operator can viewdiagnostics and results in real time, promptly respond if analyticalissues arise, and also optimize discrete sampling.

RESULTS AND DISCUSSIONTheory. Dissolved gases in the seawater flowing through the

equilibrator cartridge equilibrate with the headspace through themicroporous membrane contactor. In the absence of a capillaryor with a small capillary flow, the headspace gases approachequilibrium with the flowing seawater following Henry’s law. Atsteady state, the concentration gradient of dissolved gases betweenintake and outflow of water passing through the equilibrator isproportional to the net flux of gases across the membranecontactor. Headspace concentrations vary as a function of the netflux of gases across the membrane and the gas efflux associatedwith capillary flow to the mass spectrometer. At steady state, thenet flux of gases across the membrane from seawater to theheadspace must equal the capillary drawdown. As the capillarydrawdown of gases from the headspace increases, the gases inthe headspace become increasingly undersaturated relative totheir predicted equilibrium partial pressures. To quantify thedisequilibrium, we present a simple model that accounts for thekinetics of gas transfer across the membrane contactor and itsdependence on various parameters such as seawater flow rate andheadspace volume.

1. Kinetics of Equilibration. The gas mass balance on thewater side of the equilibrator is

VwdCdt

)Qw(Cin -Cout)-KwA(C-RChs) (1)

where Kw, Vw, Qw, A, and R are the liquid-phase mass transfercoefficient, volume of the water side of the equilibrator, theseawater flow rate, the surface area of the membrane contactor,and the Ostwald solubility coefficient (dimensionless) for thegiven gas, respectively (see Table 1 for summary of parametersused in the model). R is in effect the reciprocal of HL, thedimensionless Henry’s law constant.20 C, Cin, Cout, and Chs arethe molar concentrations within the equilibrator on the waterside, flowing in and out of the equilibrator, and within theheadspace of the equilibrator, respectively. The term on theleft-hand side of eq 1 represents the rate of change of the standingstock of a given dissolved gas in the water flowing through the

(18) Williams, P. J. L.; Jenkinson, N. W. Limnol. Oceanogr. 1982, 27, 576–584.(19) Winkler, L. W. Ber. Deutsch. Chem. Ges. 1888, 21, 2843–2855.

(20) Schumpe, A.; Quicker, G.; Deckwer, W. D. Adv. Biochem. Eng. 1982, 24,1–38.

Table 1. Parameters Used in the Model, withDefinitions and Units

parameter identification unitsQw volumetric seawater flow rate L s-1

Qc volumetric capillary flow rate out of theheadspace to the massspectrometer at headspace pressure

L s-1

F sum of the volumetric efflux ratesfrom the headspace

L s-1

Vw volume on wet side of equilibrator LVhs volume on headspace of equilibrator LC average dissolved gas molarity within

water side of equilibratorµmol L-1

Cin dissolved gas molarity enteringequilibrator

µmol L-1

Cout dissolved gas molarity exitingequilibrator

µmol L-1

Chs gas concentration in the headspaceof equilibrator

µmol L-1

A surface area of the membranecontactor

m2

R Ostwald solubility coefficient dimensionlessε equilibration coefficient dimensionlessP pressure within the headspace of

equilibratorPa

L length of capillary mr radius of capillary mη gas dynamic viscosity Pa · s (or kg m-1 s-1)Kw liquid-phase mass transfer coefficient m s-1

τ e-folding response time of equilibrator s

1857Analytical Chemistry, Vol. 81, No. 5, March 1, 2009

equilibrator. The first term on the right-hand side of eq 1represents the net flux of a given dissolved gas as water flowsthrough the equilibrator, whereas the second term equals the netflux across the membrane. The headspace concentration isassumed to be uniform, whereas the water side of the equilibratorhas a concentration gradient from the inlet to the outlet. We alsofirst assume that the net gas exchange across the membranecontactor of the equilibrator is proportional to the concentrationgradient between the seawater and the headspace (i.e., thestagnant boundary layer model21).

Likewise, the mass balance on the headspace side of theequilibrator is

Vhs

dChs

dt)KwA(C-RChs)-QcChs (2)

where Qc is the capillary flow rate (i.e., volume of gas flowingout of the headspace at headspace pressure). As in eq 1, theterm on the left-hand side of the equation represents the rate ofchange of the standing stock of a given gas. The second term onthe right-hand side of the equation is the molar flow rate of a

Figure 2. Time-series of EIMS 2 min averaged shipboard results during the Southern Ocean GasEx cruise on the NOAA ship Ronald H.Brown on April 4-5, 2008. Measurements presented in the figure were performed in the Western South Atlantic, from 51.46° S, 37.33° W to49.63° S, 40.43° W. (a) Pressure (mbar) in the quadrupole mass spectrometer (QMS), (b) percent total O2 saturation (with preliminary optodecalibration) and EIMS biological supersaturation (primary ordinate, in blue and black, respectively) and O2/Ar ion current ratio from MSmeasurements (secondary ordinate, in gray), (c) N2/Ar ion current ratio, (d) optode measurement of seawater temperature (°C) at EIMS inlet,and (e) seawater flow rate through the equilibrator (mL min-1). The red portions of the signals in panels a-c represent air calibrations. Themagenta and cyan circles in panel b represent discrete isotope ratio mass spectrometry (IRMS) O2/Ar and Winkler O2 measurements, respectively.In this example, the O2/Ar ion current ratio of air calibrations was 22.54 ( 0.01 (error bounds represent the standard deviation, n ) 12 over aperiod of 24 h). The O2/Ar saturation is calculated from the O2/Ar ion current ratio divided by the linearly interpolated O2/Ar ion current ratios ofbounding air calibrations. Flow was stopped and the optode was removed from water at time 00:00, which explains the temperature (panel d)and flow rate (panel e) signal excursions at that time. Achieved accuracy for field dissolved O2/Ar measurements by EIMS is (0.3% (standarddeviation). On this cruise, no offset in O2/Ar was observed between discrete samples collected from Niskin bottles and from the ship’s underwaysystem.


given gas down the capillary to the mass spectrometer. We assumethe capillary flow rate to be

Qc ≈ 103( πr4P16Lη) (3)

where r and L are the radius and length of the capillary (m). P isthe absolute pressure (Pa) of the headspace gas, and η is the gasdynamic viscosity (in (Pa · s) or (kg m-1 s-1)). The 103 multiplieris a unit conversion factor from cubic meters to liters. Equation3 is a simplification of the fluid dynamic law of Hagen-Poiseuille22

for viscous compressible fluids with negligible pressure at theoutlet. The viscosities of Ar, O2, and N2 at 300 K are about 2.27× 10-5, 2.06 × 10-5, and 1.79 × 10-5 kg m-1 s-1, respectively.23

The viscosity of standard air at the same temperature isapproximately 1.87 × 10-5 kg m-1 s-1. Under our experimentalconditions, the pressure in the headspace of the equilibratoris close to 105 Pa (1 atm). On the basis of the capillary’sdimensions, the volumetric flow rate to the mass spectrometershould be around 0.2 µL s-1 or 18 mL day-1 (standardtemperature and pressure). We measured the flow rate to be19 mL day-1 by determining the decrease in pressure of anisolated gas volume as its air was allowed to flow through acapillary to vacuum. Hence, our simplified equation seems toprovide a reasonable estimate of capillary flow for the condi-tions used during these experiments.

We neglect the slight decrease in headspace pressure thatcomes from the fact that the gas flux down the capillary issignificant relative to the gross flux across the membranecontactor into the headspace. As the capillary flow increasesrelative to the gas fluxes across the membrane (e.g., shortercapillary), the pressure decreases within the headspace. Thisdecrease in pressure reduces the capillary flow rate (see eq 3)and increases the net flux across the membrane, partly compen-sating for the effect of the capillary flow. Our calculations of thedegree of disequilibrium should therefore be regarded as upperlimits. Depending on the configuration, the pressure in theheadspace of the equilibrator is within ±10% of 1 atm.

Assuming that the change with time of the standing stock ofa given gas on the water side of the equilibrator is negligible,and assuming that the dissolved gas exiting the equilibrator hasfully equilibrated with the headspace, and rearranging eqs 1 and2, the change in the headspace standing stock of a given gas canbe approximated as

Vhs

dChs

dt)QwCin - (QwR+Qc)Chs (4)

Defining F ) (QwR + Qc) (i.e., the sum of the volumetricefflux rates from the headspace) and integrating relative to time(i.e., time after change in gas concentration of seawaterentering equilibrator)

Chs(t)) 1

F[(FChs0

-QwCin)e-(F/Vhs)t +QwCin] (5)

where Vhs/F is the e-folding response time (τ) of the equilibra-tor. The transit time for gases through the capillary is on theorder of 10 s and hence is short compared to the responsetime associated with the headspace equilibration. If there isno flux of gas down the capillary, this equation simplifies tothe model presented by Johnson24 for a showerhead CO2

equilibrator.The e-folding residence time calculated based on the above

equation (approximately 0.7 min) is shorter than our empiricalobservations (see results below), which suggests that the assump-tion in eq 4 that the seawater flowing out of the equilibrator hasfully equilibrated with the headspace is invalid. In order to accountfor the relatively slow exchange of gases across the membrane,a dimensionless equilibration coefficient (ε) between 0 and 1(1 being a full equilibration) may be applied24 to the gas exchangeacross the membrane. Taking into account the kinetics of gastransfer across the membrane, eq 5 becomes

Chs(t)) 1

F[(FChs0

- εQwCin)e-(F/Vhs)t + εQwCin] (6)

where F now equals (εQwR + Qc). We assume that most of theresistance to mass transfer is from the aqueous phase (Kw)and is mediated by molecular diffusion on the water side ofthe membrane. Kw is then a function of the molecular diffusivitycoefficients of the gases. Differences in diffusivity betweengases can be accounted for by normalizing the equilibrationcoefficients of the various gases to their relative moleculardiffusivity coefficients. The molecular diffusivity coefficients ofN2, O2, and Ar, as compiled by Broecker and Peng,25 are 2.1 ×10-5, 2.3 × 10-5, and 1.5 × 10-5 cm2 s-1 at 24 °C, respectively.However, the coefficients reported in the literature varysubstantially.26-28

A low equilibration coefficient would be consistent withsignificant resistance to the gas mass transfer across the mem-brane from the aqueous boundary layer, the air-filled pores(hydrophobic material) of the membrane, and the gas boundarylayer in the headspace. The reciprocal of the overall mass transfercoefficient, known as the resistance, is a function of the resistancefrom these various mediums. Yang and Cussler29 expressed theoverall resistance with a series model analogous to Ohm’s lawfor electrical resistors in series where the overall mass transfercoefficient (K) is a function of the aqueous medium, the mi-croporous membrane, and the headspace transfer coefficients. Inmost applications, a sweep gas is applied to the other side of themembrane contactor. Under such conditions, the resistance tomass transfer across the hollow fiber contactor is generallydominated by the aqueous phase.29 This is because moleculardiffusion of gases in water is orders of magnitude slower than inair. However, under certain conditions, such as low flow rate ofthe sweep gas relative to the water flow rate, the air-filled poresand the headspace may represent significant resistances to the

(21) Lewis, W. K.; Whitman, W. G. Ind. Eng. Chem. 1924, 16, 1215–1220.(22) Golubev, I. F. Viscosity of Gases and Gas Mixtures; Fizmat Press: Moscow,

1959.(23) Lemmon, E. W.; Jacobsen, R. T. Int. J. Thermophys. 2004, 25, 21–69.

(24) Johnson, J. E. Anal. Chim. Acta 1999, 395, 119–132.(25) Broecker, W. S.; Peng, T. H. Tracers in the Sea; Eldigio Press: New York,

1982.(26) Baird, M. H. I.; Davidson, J. F. Chem. Eng. Sci. 1962, 17, 87–93.(27) Wise, D. L.; Houghton, G. Chem. Eng. Sci. 1966, 21, 999–1010.(28) Ferrell, R. T.; Himmelblau, D. M. J. Chem. Eng. Data 1967, 12, 111–115.(29) Yang, M. C.; Cussler, E. L. AIChE J. 1986, 32, 1910–1916.


mass transfer of gases.30,31 The headspace of our equilibrator iseffectively stagnant. Hence, the membrane and the headspace mayalso represent a substantial resistance to the mass transfer of gasesin EIMS.

2. Steady-State Conditions. At steady state (i.e., tf∞), Chs

tends toward a concentration that is proportional to the ratioof the volumetric influx and efflux from the headspace:

Chs(tf ∞))

εQw

(εQwR+Qc)Cin (7)

If the capillary flow is small relative to the gas exchange acrossthe membrane, the gas concentration in the headspace tendstoward gas equilibration, i.e., Cin/R. The steady-state ratio of twogases (gas 1 and gas 2) in the headspace is therefore (Figure3, parts A and C)

Chsgas1

Chsgas2

) (Cingas1

Cingas2)[ (Rgas2 +Qc(εgas2Qw)-1)

(Rgas1 +Qc(εgas1Qw)-1)] (8)

For two gases with similar solubilities and molecular diffusivi-ties, the capillary drawdown of gases does not significantly changethe steady-state headspace gas ratio from equilibrium gas ratiopredictions (Figure 3B). This is because the numerator anddenominator within brackets in eq 8 in effect cancel out. On theother hand, for two gases such as N2 and Ar, with very differentsolubilities, an increase in capillary flow relative to the gasequilibration rate increases the calculated disequilibrium in theheadspace gas ratio with the flowing seawater (Figure 3, partsC and D). To approach conditions of a fully equilibrated head-space, more than twice as much N2 as Ar must cross themembrane relative to their concentrations in seawater. The fluxof gas to the mass spectrometer depletes headspace concentra-tions. The flux of gases across the membrane cannot quitemaintain gases within the headspace at the partial pressuresin seawater. The lower the solubility of a gas, the greater is itspartial pressure deficiency relative to equilibrium in theheadspace. N2, which has a lower solubility than Ar, is thusmore deficient in the headspace, accounting for the observedreduction in N2/Ar ratio of headspace gases compared withfull equilibration (Figure 3, parts C and D). The effect is smallerfor the O2/Ar ratio (Figure 3B) because these two gases havesimilar solubilities. However, it is possible to observe an O2/Ar

(30) McDermott, C. I.; Tarafder, S. A.; Kolditz, O.; Schuth, C. J. Membr. Sci.2007, 292, 17–28.

(31) Tarafder, S. A.; McDermott, C.; Schuth, C. J. Membr. Sci. 2007, 292, 9–16.

Figure 3. Modeled relaxation of O2/Ar (A and B) and N2/Ar (C and D) in the headspace as a function of seawater flow rate (A and C) andcapillary length (B and D). Equilibration coefficient (ε) of O2 is assumed to be 1 (seawater flowing out is fully equilibrated with the headspace).Model is based on eq 6 with an equilibration coefficient (ε) for N2 and Ar normalized to the molecular diffusivity of each gas relative to O2 at 24°C. The molecular diffusivity coefficients of N2, O2, and Ar used in this example are 2.1 × 10-5, 2.3 × 10-5, and 1.5 × 10-5 cm2 s-1, respectively(ref 25). The difference in diffusivity of oxygen and argon may, however, be smaller (ref 27), in which case, the steady-state O2/Ar supersaturationpresented in this figure would be overestimated. The headspace gas ratios are 10% supersaturated in O2/Ar and N2/Ar at time 0, at which pointwater flow is switched to air-equilibrated seawater. The varying seawater flow rate (A and C) is modeled with a 2 m long capillary. The varyingcapillary length (B and D) is modeled with 100 mL min-1 seawater flow rate through the equilibrator.


supersaturation in the headspace if the oxygen has a greatermolecular diffusivity than argon25 and if the drawdown of gasesfrom the headspace is significant relative to the gas flux acrossthe membrane (Figure 3B). The effect is, however, smaller thanfor N2/Ar (Figure 3). On the basis of eqs 7 and 8, it can be shownthat the level of undersaturation relative to equilibrium for a givengas in the headspace is proportional to the ratio of capillarydrawdown to gross flux across the membrane into the headspace.Similarly, the under- or supersaturation relative to equilibrium inelemental ratio of two gases in the headspace is proportional tothe difference of the ratios of capillary flows to gross fluxes intothe headspace for the two gases measured. In our currentconfiguration, we estimate the capillary flux to be roughly 1% ofthe gross flux across the membrane into the headspace foroxygen.

For accurate EIMS measurements of gases with differentsolubilities, such as N2/Ar, the capillary flow must be negligiblerelative to the gas equilibration rate. Alternatively, a correctionto the N2/Ar measurements could potentially be applied,assuming the seawater flow rate, gas flow across the membrane(boundary layer conditions), and capillary gas drawdown fromthe headspace are well constrained by measurements or theoryand validated by calibrations. Because boundary layer condi-tions are affected by many factors such as changes in temper-ature and flow rate, such a correction may be difficult to applywhen capillary flow is significant relative to the gas flux acrossthe membrane.

Laboratory Tests of EIMS. 1. O2/Ar: Kinetics of Equili-bration. The response time of EIMS to a change in gasconcentrations, as well as its linearity, were tested in the laboratoryby switching the source of water flowing through the equilibratorfrom air-equilibrated water (i.e., dissolved gas pressure in equi-librium with atmospheric gas pressure) to waters with elevatedO2 concentrations (Figure 4A). O2 supersaturated waters, withO2/Ar ratios elevated above saturation ratios by 2.2%, 10.1%,and 12.4% were prepared by adding different amounts of oxygento the headspace of Tedlar bags containing air-equilibratedwater. The e-folding response time (τ) (i.e., t1/2/ln(2)) wasdetermined by fitting the O2/Ar ion current ratio I(m/z ) 32)/I(m/z ) 40) (denoted 32/40) to a reaction progress equationbased on eq 6

(3240)(t)

) [(3240)i

- (3240)f]e(-t/τ) + (32

40)f(9)

where (32/40)i and (32/40)f are the initial and final 32/40,respectively. τ is estimated as the reciprocal of the slope of alinearized form of eq 9 with time as the independent variable

y) ln[ (3240)(t)

- (3240)f

(3240)(i)

- (3240)f

] )-(1τ)t (10)

Kinetics of equilibration were estimated from the response ofthe 32/40 as we switched the flow through the equilibratorcartridge between air-equilibrated water and O2/Ar-elevatedwaters. From these experiments, the e-folding response timeis found to be on average 7.75 ± 0.25 min (± standard deviation)

(Figure 4B). On the basis of these results, a change in the O2/Ar ratio of 3% will be measured with a bias (i.e., memory effect)of 0.06% after 30 min of equilibration. The response is slowerthan predicted based on the model presented in eq 5. Observa-tions can be reconciled with the model by invoking an equilibra-tion factor of 0.1 (ε in eq 6), implying that the kinetics of gastransfer are relatively slow.

O2/Ar: Steady-State Conditions. The linearity of theresponse was determined by comparing the relative change insignal on the EIMS to values measured independently in

Figure 4. EIMS O2/Ar response time and linearity when switchingfrom equilibrated water to waters with various O2/Ar supersaturations.Experiments were performed with a seawater flow rate of 100 mLmin-1. O2/Ar supersaturated waters were prepared by injecting 10(green), 50 (red), or 100 mL (blue) of pure O2 to the small headspaceof 3 L of equilibrated water contained in Tedlar bags. (A) EIMS O2/Arion current ratio (32/40) as a function of relative time for the variousO2 injections. The three experiments were performed on differentdays. (B) y ) ln[((32/40)(t) - (32/40)f)/((32/40)(i) - (32/40)f)] ) -(1/τ)t (fromeq 10) for the red and blue experiments. The upper blue and redcurves (1 and 3) are for the kinetics of perturbation of O2/Arequilibrium, whereas the lower blue and red curves (2 and 4) are forthe kinetics of relaxation to O2/Ar equilibrium. The time of onset ofperturbation and relaxation has been normalized to zero. For clarity,the intercept of each curve has been modified to avoid overlappingof curves. The dashed lines represent linear least-squares values ofthe slopes. From top to bottom, the e-folding times, the reciprocalsof the slopes, are 7.80 (r2 ) 0.98), 7.41 (r2 ) 0.97), 8.01 (r2 ) 0.98),and 7.78 (r2 ) 0.97) min. (C) Percent deviation from equilibrium inthe O2/Ar ratio of waters with various O2/Ar mixtures as measuredon discrete samples by isotope ratio mass spectrometry (abscissa)and as measured on the EIMS (ordinate). Each data point on thegraph represents one switchover from air-equilibrated water to waterwith a different O2/Ar mixture. The black filled circle at the originrepresents comparisons of equilibrated waters vs ambient air (Figure5, where the mean difference in %O2/Ar between equilibrated waterand ambient air is less than 0.1%). The slope of the comparison forthe colored points is not significantly different from the identity slope,represented by the black line (t value ) 0.94, p ) 0.48, DF ) 1).


discrete samples analyzed by isotope ratio mass spectrometry(IRMS), as described by Emerson et al.32 The two methodsmeasure similar fractional changes when switching amongsolutions of different O2/Ar ratios (Figure 4C).

The linearity of the response is further supported by fieldobservations. O2/Ar saturation measurements, derived from 5min time-averaged O2/Ar ion current ratios and air calibrations,are within 0.3% (standard deviation) of corresponding IRMSmeasurements on discrete samples (mean difference ) 0.07%,n ) 75, example shown in Figure 2). This is similar to theprecision achievable by MIMS.14 In addition, under circumstanceswhere variability in total oxygen concentration in the surface oceanis driven by biological activity, the total and the biological oxygensupersaturations are expected to covary. This statement is true ifall the oxygen supersaturation within the surface of the ocean isderived from NCP or if physical processes such as bubble injectionimpart a constant oxygen supersaturation to the biological signal.In line with this expectation, we find that changes in O2/Ar ioncurrent ratio closely track changes in percent O2 saturation asobserved by the optode (Figure 2B). This observation suggeststhat, over the time scale of a day or so, most variability in the32/40 ion current ratio derives from changes in biological O2

supersaturation and that the 32/40 ratio measured with EIMSresponds linearly to the seawater ratio. The unlikely alternative

is that the physical supersaturation of Ar exactly compensatesfor nonlinearities in the mass spectrometer.

Although the slow kinetics of gas transfer across the mem-brane contactor affect the equilibration rate of the O2/Ar ratio,they do not significantly fractionate the O2/Ar once steady stateis achieved. This conclusion is validated by an experiment inwhich we switched between a capillary sampling headspacegas from an equilibrator through which air-equilibrated waterflowed and capillaries of the same length sampling ambientair (Figure 5). It should, however, be noted that, if the moleculardiffusivities of O2 and Ar in fact differ,25 a small but significantdeparture from equilibrium O2/Ar at steady state may occur ifthe ratio of capillary to gas flux across the membrane is greaterthan in our current configuration.

To test the dependency of steady-state O2/Ar measurementson seawater flow rate, we measured the O2/Ar ion currentratios as we varied the flow of water through the equilibrator.No discernible differences in the mean O2/Ar ion current ratios(n ) 150) were observed at flow rates of 90, 110, and 135 mLmin-1 (mean (32/40) of 25.64 for each flow rate).

O2/Ar: Air Calibrations of O2/Ar Ion Current Ratio. Inorder to test ambient-air calibrations, we examined the changein the O2/Ar ratio when switching from air-equilibrated water(sampled from the headspace of the equilibrator cartridge) toambient air (see above). This change was very small. Becauseof potential drift in the instrument, the mean O2/Ar ion currentratio of equilibrated water and dry air were calculated only fromthe 200 measurement cycles immediately preceding and fol-lowing the switch from equilibrated water to dry air. Sixcomparisons were performed (see Figure 5). A 95% confidenceinterval for the true difference in mean ion current ratio betweenair-equilibrated water and dry air shows that, although thedifference in O2/Ar ion current ratio is significant, the discrep-ancy between the two measurements is small: 0.04-0.12%. Ourmeasurements therefore show that the O2/Ar partial pressureratio measured with EIMS is negligibly fractionated betweenwater and headspace in our current configuration. For com-parison, the achievable precision with the standard IRMSmethod on discrete samples is ±0.1-0.35%.11,32

Water vapor decreases the mass spectrometer’s ionizationefficiency and reacts in the mass spectrometer to form O2.33 Inthe EIMS configuration, the headspace of the equilibrator iswater-saturated, whereas the water vapor pressure of thecalibration gas, ambient air, is variable. To test the effect ofvarying water partial pressure, the 32/40 ion current ratio ofwater-saturated air (i.e., headspace air of a sealed containerwith water) and dry air (i.e., air in a container with somecalcium sulfate desiccant) were sequentially measured. Sixcomparisons of means of dry and water-saturated air ion currentratios were performed by sequential EIMS measurements ofdry and water-saturated air (Figure 5). The differences in meanO2/Ar ion current ratios of water-saturated and dry air werenot statistically significant (p > 0.05, DF ) 398) (Figure 5A).

A variety of processes have the potential to alter the dissolvedO2 concentration of seawater as it is pumped through the ship’splumbing into the laboratory.14 These include heterotrophic

(32) Emerson, S.; Stump, C.; Wilbur, D.; Quay, P. Mar. Chem. 1999, 64, 337–347.

(33) Orsnes, H.; Bohatka, S.; Degn, H. Rapid Commun. Mass Spectrom. 1997,11, 1736–1738.

Figure 5. Laboratory assessment of air calibrations of ion currentratios. (A) O2/Ar ion current ratio (32/40). (B) N2/Ar ion current ratio(28/40). (C) Pressure recorded in the mass spectrometer. Interfacedwith a Valco valve, the quadrupole mass spectrometer sequentiallymeasures (1) air from the equilibrator through a 480 mm capillary(blue), (2) ambient dry through a 480 mm capillary (red), (3) ambientwet through a 480 mm capillary (green), (4) ambient dry air througha 580 mm capillary (black), and (5) ambient dry air through a 380mm capillary (cyan). Air-equilibrated water is flowing through theequilibrator and the headspace gas is measured in (1) (blue).


activity within the lines, rust, warming and degassing, andcavitation. Hence, discrete sampling independent of the ship’sseawater underway sampling system (e.g., Niskin bottles) isneeded to correct for any O2 depletion within the ship’s linesand within the EIMS lines upstream of the equilibrator.

2. Preliminary Test of N2/Ar Measurements by EIMS.The N2/Ar concentration ratio in normal air is approximately83.6 compared with 37.9 in air-equilibrated seawater (S ) 35,T ) 20 °C). The dissolved N2/Ar naturally varies in the surfaceoceans predominantly because of bubble injection and tem-perature changes.13 Cooling and bubble injection cause N2/Ar in the mixed layer to exceed the saturation value. The slowkinetics of gas exchange across the membrane contactor,together with the significant drawdown of headspace gasesfrom the capillary, may explain the observed undersaturatedN2/Ar gas ratio observations (Figure 5). When the capillarydrawdown of headspace gases is stopped (e.g., switch fromequilibrator measurements to air calibrations depicted in Figure5), the N2/Ar in the headspace increases due to the replenish-ment of N2 and Ar from gas exchange across the membrane.As the capillary flow is reestablished (e.g., switch from aircalibrations to equilibrator measurements depicted in Figure5), the N2/Ar in the headspace starts out very close to theequilibrium but then decreases by about 1%.

An alternative explanation for the anomalous 28/40 ion currentratio might be fractionation in the source due to the difference insource pressures between sample and air analyses. However, the28/40 ion current ratio actually rises rather than falls as flow rateto the mass spectrometer decreases. This result is shown in Figure5: with the capillary sampling ambient air, the shorter the capillary,the greater the pressure in the mass spectrometer, and the lowerthe 28/40 signal (Figure 5C). In our field observations (Figure2), the departure from equilibrium due to the difference insolubility between N2 and Ar is exacerbated by the slightpressurization of the headspace.

3. Preliminary Test of CO2 Measurements by EIMS. TheCO2 in the equilibrated headspace of the EIMS should bereflective of the dissolved CO2 partial pressure. Initial fieldcomparison of EIMS ion current ratio 44/40 (i.e., CO2/Ar) toCO2 measured by the standard showerhead equilibrator inletnondispersive infrared (NDIR) analyzer were performed on theAustralian icebreaker Aurora Australis in January-February2007 in the Australian subantarctic zone. Because the (44/40)ion current ratio may drift over time, we compared EIMSmeasurements with NDIR CO2 estimates over 6 h intervals.The median 6 h Pearson correlation coefficient between thetwo measurements for the last 10 days of the cruise was 0.97.(44/40) measurements divided by independent NDIR pCO2

estimates scatter about a linear trend by 0.2% (<0.8 ppm CO2)in a single 24 h period (Figure 6). During these preliminaryexperiments, we did not calibrate the mass 44 signal with CO2

standards, which explains the drift between the two measure-ments (Figure 6). These preliminary field observations nonethe-less suggest that pCO2 measurements by EIMS are feasible andpotentially offer precision comparable with the current standardmethod. High-resolution measurements of CO2 by MIMS havealso recently been performed.34

Advantages and Disadvantages of EIMS Relative to Mem-brane Inlet Mass Spectrometry. As opposed to EIMS, the inlet(i.e., space opposite of the water side of the membrane) in a MIMSconfiguration is under a vacuum.15 Under such conditions, thediffusion of gases across the membrane back to the seawater isnegligible relative to the gas flow to the mass spectrometer.Hence, in the MIMS configuration, the molar flow rate acrossthe membrane at steady state simplifies to KAC(see eq 2), andthe flow to the mass spectrometer is the same. Under suchconditions, the molar flow rate to the mass spectrometer isobviously a function of the membrane’s permeability (here K) toa given gas. The MIMS e-folding response time is a function ofthe total volume on the vacuum side of the membrane divided bythe flow of gas to the mass spectrometer. Because volume on thevacuum side is small, a significant advantage of MIMS is the fasterresponse time relative to EIMS. In MIMS, the response time isin effect instantaneous (i.e., seconds). In the EIMS configuration,the response time is dependent on the equilibrator’s headspaceturnover time (see model above). In regions with high spatialheterogeneity such as high-productivity coastal ecosystems, thefaster response time of the MIMS may be desirable. On the otherhand, the response time of EIMS is sufficient to capture most ofthe variability in open ocean waters where spatial gradients areless compressed. The faster response time of the MIMS alsomeans that discrete samples from depth profiles can be measuredon the MIMS.16 Such measurements would be impractical onEIMS because large volumes and sampling times would beneeded.

Because of the vacuum, the membranes in MIMS applicationsare generally nonporous.15,16 Under such conditions, the gasesmust diffuse through the membrane material. The gas fluxes tothe mass spectrometer are therefore dependent on the mem-brane’s selective permeability to the various gases.35 Because thegases can be stripped out of solution and because membraneswith selective permeability can be used, the measurement of tracedissolved gases is more efficient in MIMS than in EIMS. In EIMS,(34) Gueguen, C.; Tortell, P. D. Mar. Chem. 2008, 108, 184–194.

Figure 6. EIMS CO2/Ar ion current ratio (44/40) and pCO2 at ambienttemperature as determined with an underway nondispersive infraredanalyzer (LI-COR) showerhead equilibrator during the SAZ-Sensecruise in 2007 onboard the ship Aurora Australis in the Australiansector of the Southern Ocean (at approximately 54.2° S, 146.4° E).The EIMS CO2/Ar has not been calibrated with CO2 standards, whichexplains the drift relative to the other method.


the equilibration occurs mostly through the micropores of themembrane. Hence, the gas ratios on the headspace side in EIMSare dictated by the gases’ solubility coefficients. The partialpressure of a given gas in the headspace of the equilibrator willbe similar to the partial pressure in solution. Determination oftrace gases in seawater may therefore be difficult or impossibleby EIMS. For example, variations in dimethyl sulfide (DMS), avolatile organic compound released by phytoplankton and aputatively important control on global climate (CLAW hypoth-esis36), is measurable by MIMS37 but may not be measurable byEIMS.

In MIMS, the membrane permeability is affected by itsdiffusive boundary layer thickness, which is dependent on thewater flow rate, as well as other factors including turbulence,temperature, and bubbles within the system.15,16 The lowersensitivity in EIMS to changes in boundary layer thickness maybe due to the fact that both fluxes (i.e., influx and efflux) acrossthe membrane are equally affected by changes in boundary layerconditions. The longer residence time in EIMS also potentiallybuffers short-term variations in boundary layer conditions. Thisreasoning does not apply when capillary flow is significant relativeto gas flux across the membrane, in which case boundary layercondition changes may affect the measurements.

For the reasons mentioned above (most importantly themembrane’s selective permeability to various gases), calibrationof the MIMS must be performed with air-equilibrated waters ofknown temperature and salinity. Calibration of EIMS O2/Ar canbe performed with ambient air. The MIMS requires activetemperature control for membrane and equilibrated waters,15,16

more space, more daily attention, and its installation is more

complex. In the case of EIMS, temperature is maintained at theseawater value using a high seawater flow rate to thermostat theequilibrator cartridge (Figure 1). Depending on the application,EIMS or MIMS measurements may be more suitable.

CONCLUSIONBiogeochemical properties in the surface ocean respond to

mesoscale and submesoscale variability in ocean physics. Theseproperties may therefore vary at high frequencies in space andtime. Because of differences in time scales of biogeochemical andphysical properties, spatial heterogeneity in the former mayexceed the later.38 Hence, extrapolation of biogeochemical ob-servations based on physical property homogeneity requirescaution. Incidentally, current observations often fail to capture thelarge variability in ocean productivity.39,40 We have presented anew method for continuous, high-resolution measurements ofdissolved O2/Ar. From the continuous O2/Ar measurementsby EIMS, NCP can be resolved at high spatial resolution. Thesehigh-resolution measurements will provide new insight into thebiogeochemical cycling of carbon in the ocean’s mixed layer.

ACKNOWLEDGMENTWe are grateful to Peter DiFiore (Princeton, U.S.A.) for

operating the EIMS during the SAZ-Sense cruise in the SouthernOcean. We also thank Ralph Keeling (Scripps, U.S.A.) and HiroakiYamagishi (NIES, Japan) for helpful discussions. This researchwas supported by funding from NASA, NSF, and the Princeton-BPAmoco Carbon Mitigation Initiative.

Received for review October 31, 2008. Accepted January9, 2009.

AC802300U(35) Reed, B. W.; Semmens, M. J.; Cussler, E. L. In Membrane Separations

Technology, Principles and Applications; Noble, R. D., Stern, S. A., Eds.;Elsevier Science: Amsterdam, The Netherlands, 1995; p 738.

(36) Charlson, R. J.; Lovelock, J. E.; Andreae, M. O.; Warren, S. G. Nature 1987,326, 655–661.

(37) Tortell, P. D. Geochem. Geophys. Geosyst. 2005, 6, Q11M04.

(38) Mahadevan, A.; Campbell, J. W. Geophys. Res. Lett. 2002, 29, 1926.(39) Williams, P. J. L.; Morris, P. J.; Karl, D. M. Deep-Sea Res., Part I 2004, 51,

1563–1578.(40) Karl, D. M.; Laws, E. A.; Morris, P.; Williams, P. J. L.; Emerson, S. Nature

2003, 426, 32–32.


Documents

Continuous High-Frequency Dissolved O 2 /Ar Measurements by Equilibrator Inlet Mass Spectrometry