Metered reagent injection into microfluidic continuous flow sampling for conductimetric ocean dissolved inorganic carbon sensing

Continuous and autonomous measurement of total dissolved inorganic carbon (TCO2) in the oceans is critical for modelling important climate change factors such as ocean uptake of atmospheric CO2 and ocean acidification. Miniaturised chemical analysis systems are therefore required which are small enough for integration into the existing Argo ocean float network for long-term unattended depth profiling of dissolved CO2 with the accuracy of laboratory bench analysers. A microfluidic conductivity-based approach offers the potential for such miniaturisation. Reagent payload for>3 yr operation is a critical parameter. The precise injection of acid into sample, liberating CO2 from seawater, is addressed here. Laser etched microfluidic snake channel restrictors and asymmetric Y meters were fabricated to adjust the metering ratio between seawater and acid simulants. Laser etching conditions were varied to create a range of channel dimensions down to ~75 microns. Channel flow versus pressure measurements were used to determine hydrodynamic resistances which were compared with finite element simulations using a range of cross-section profiles and areas. Microfluidic metering circuits were constructed from variable resistance snake channels and dimensionally symmetric or asymmetric Y-junctions. Sample to acid volume ratios (meter ratio) up to 100:1 have been achieved with 300 microns wide snake channel for lengths>1m. At the highest pattern resolution, this would require a footprint of>600 mm2 (6 x10-4 m2). Circuits based solely on asymmetric Y-junctions gave meter ratios up to 16:1 with a footprint cost of<40 mm2 and precision values of ~0.2%. Further design and fabrication refinements will be required to ensure the structural and dimensional integrity of such small channels in future integration of metering units into full TCO2 analysis microfluidic circuits.


I o uc o
Measurement of tota disso ved O 2 (T O 2 ) content in seawater by miniaturised sensors in deep sea f oats is set to become important for ong term oceanic monitoring as the oceans capture the increasing amounts of O 2 re eased by fossi fue burning [ ]. Such measurements wi feed into c imate change mode ing to improve the accuracy of future g oba temperature rise predictions as monitored by the [2] and sea eve rise prediction [3]. ew microf uidic ab on chip devices for T O 2 measurement must be deve oped for ong-term autonomous dep oyment in a deep ocean environment. This poses severe cha enges with regard to design fabrication and testing. Oceanic studies of T O 2 content in seawater have traditiona y used samp e co ection by research vesse s and aboratory ana ysis. This is s ow expensive and has provided on y sparse coverage whereas continuous remote monitoring across the wor d's oceans and to depths of up to 3km are required. Severa methods have been demonstrated for fie d ana ysis of T O 2 such as conductivity change [4 -6] spectrophotometry [7 -0] R absorption ana ysis [ -4] gas chromatography [ 5] and membrane in et mass spectroscopy [ 6 -8]. Partia pressure of surface O 2 p O 2 has been measured from buoys f oats and ships [ 9 -2 ]. Microf uidics based conductivity and temperature sensors have been tria ed in the ocean [25] with other devices a so investigated for oceanic nitrate/nitrite and phosphate/phosphorus detection [28 -3 ] ammonium ion measurement [32] manganese sensing [33] ocean acidification [34] biochemica and microbio ogy app ications in simi ar extreme environments [26 27].
The Argo network of ~5000 untethered autonomous ocean f oats offers the possibi ity of continuous ow cost T O 2 measurements to the required spatia and depth reso ution and therefore inspires the deve opment of a suitab e T O 2 microf uidic p atform which cou d meet the stringent constraints on size and power as we as re iabi ity in the harsh ocean environment. Argo f oats are current y dep oyed to measure pressure temperature and sa inity using rea time (instantaneous) samp ing and measurement as they ascend to the surface from their target depth of 2 -3 km [39]. The f oats then transmit the data to sate ite descends to a drift depth of 500 m for 9 days before the next descent to target depth. The 0-day cyc e repeats continuous y with each untethered f oat having a projected ifetime of 3 -5 years. T O 2 measurements using aboratory prototypes or in short term ocean dep oyment have achieved the ≤ 0.2% precision thought to be necessary for ocean characterisation [5 9 0 37]. However microf uidic imp ementations which meet vo ume power cost and re iabi ity constraints strugg e to reach the required accuracy. T O 2 determination by conductivity measurement was origina y proposed by Ha and A er [4] and remains the approach with the greatest potentia for ow-cost and ow-power miniaturisation compared to optica or mass spectrometry. However with conductivity measurement noise rejection is a much more demanding cha enge. membrane exchange via diffusion prevents instantaneous measurement as the f oat ascends or descends and hence the on y option is to co ect seawater samp es in microf uidic storage ce s during ascent/descent for subsequent ana ysis whi e the f oat is parked at 500 m for 9 days. Mu ti-samp e membrane-based ana ysis poses new cha enges for microf uidics fabrication inc uding a requirement for high integrity mu ti-channe and mu ti-eve structures with robust ong-term and chemica y-resi ient bonding. n recent work we have demonstrated suitab e PMMA bonding processes for ong term mu ti-ayer and mu ti-channe operation [42] PDMS O 2 membrane sea ing within a PMMA manifo d [43] and microf uidic O 2 separation and conductivity measurement [44]. The requirement for samp e storage p us the reagent (acid aOH) pay oad for measurement and f ushing represents the primary contribution to system vo ume and maximum imits be ow ~ L wi be necessary. With P profi es (up to 50) per f oat ife and samp es per profi e (up to 00) reagent usage must u timate y be reduced to the micro itre per samp e sca e. This work investigates one such unit the injection of precise acid quantities into seawater. Reduction in seawater pH from its norma eve (~8) to ≤ 4 is possib e by adding a minimum quantity of high strength acid. To avoid the necessity of a separate acid pump which incurs a significant power and vo ume cost simu taneous acid injection during the seawater samp ing stage offers a possib e so ution. The contro ed additiona of acid is determined by the metering ratio (MR) of the microf uidic circuit where MR is the samp e to acid vo ume ratio (V S /V A ). The tota acid pay oad vo ume is therefore Px (V S /MR) and maximising MR without incurring samp e vo ume or other fabrication pena ties is the objective. For a monoprotic acid A the maximum MR at which a T O 2 is converted to O 2 is A precision of 0. % wou d require an acid concentration [A] ≥ 3M and MR ~ 000 which is not rea istic. The a ternative therefore is to inc ude the di ution factor within the device ca ibration. n this case the precision is now dependent on MR f uctuations during operation due to for examp e temperature and pump variabi ity. The maximum a owed f uctuation in MR from the origina ca ibrated va ue is shown in Fig. 2 for a specified precision of 0. %. Since the resistance of acid and seawater channe s depends on dynamic f uid viscosity the MR temperature dependence needs to be considered over the ike y ocean temperature variation from surface to 2 km depth. This variation is typica y 0 o but can reach up to 20 o . The change in MR with temperature is ca cu ated for two rectangu ar crosssection channe s Fig. 3 from the standard ana ytica approximation ( ). ( where R H is the hydrodynamic resistance of a rectangu ar channe of ength L width w and height h carrying a f uid of dynamic viscosity . The nomina MR given by R H(acid) /R H(SW) is dependent on the ratio of acid (H ) to seawater dynamic viscosity which reduces with depth as the sea temperature fa s Fig. 3. For an expected temperature at depth of 5 o and a maximum surface temperature of 25 o assuming 20% H concentration and an ocean sa inity of 0.035 kg kgthe change in MR is ~8%. To imit the effect of such a MR variation on T O 2 measurement precision imp ies a minimum MR va ue of ~75 Fig. 2. With ower H concentration the temperature dependence is attenuated. Whi e it wou d be possib e to factor the temperature dependence into the device ca ibration this wou d require simu taneous temperature measurement at each depth samp ing point.  n this paper we report the fabrication and characterisation of microf uidic structures suitab e for sing e pump use in a continuous f ow operation where metering is based on the channe resistance difference between two arms of a microf uidic junction. We imp ement various MR va ues via optimisation of channe cross-section area ( SA) or ength. The atter invo ves uniform SA which faci itates ease of fabrication at the expense of vo ume whi e the former invo ves differentia SA and presents a cha enge for fabrication due factors such as channe height oss and deformation which are inevitab e outcomes of thermop astic bonding. Microf uidic performance has a strong dependence on the minimum achievab e channe height and width va ues and their associated to erances. Whi e we have obtained sub-00 m using precision mi ing the need for ong tight y spaced channe s and variab e heights and widths within a given pattern have proven to be prob ematic with this method. nstead we use aser etching to define channe s due to the pattern f exibi ity and higher reso ution. However aser etching especia y at the finest reso ution resu ts in a oss of the traditiona rectangu ar cross-section and hence the resu tant channe resistance is no onger predictab e. A ong with pressure -f ow and meter ratio measurements we a so report a number of finite e ement simu ations with variab e cross-section profi es used to estimate channe resistance.

Exp m l
Microf uidic channe s were engraved in cast PMMA ( 0 mm thickness) using a O 2 aser ( niversa Laser Systems VLS2.30 25W source power wave ength 0.6 µm) from Auto AD patterns. A standard .5" foca ength ens was used for rastered snake channe s with engths of 330 663 994 and 320 mm a typica examp e being shown in Fig. 4. The snake channe s consisted of 30 mm straight sections with semi-circ e turns at each end. A snake channe s were rastered with settings of 00% power 25% speed (~ 0.3 m/s) and Auto AD drawing inewidth of 0.09 mm. A HPDFO (High Power Density Focussing Optics) ens was used for improved optica definition for fabrication of Y meters which required finer inewidths for narrow restrictions in a section of the acid input ine. F uidic connections were then attached via mi ed and tapped ¼-28 ports with mm through ho es. PMMA bonding was achieved via H 3 so vent vapour treatment and subsequent therma treatment (50° ≥ 20 h) to drive off residua ch oroform. Asymmetric Y meters were fabricated with raster mode channe s of nomina channe width 0.5 mm except for a thin 0 mm ong vector mode restrictive channe between the acid ine input and the Y-junction. Laser vector mode operation gives a smoother ine with straight edges in the Y direction. Four Y meters with different restrictive channe cross-sections were fabricated Fig. 5 with the Y junction axis at 45° or 225° rotation.  Representative optica cross-sections of raster and vector-written channe s in PMMA are shown in Fig. 6 and Fig. 7 respective y. The raster channe s have a f atter base because they are formed from at east 2 scan ines whi e the vector channe s are formed from one scan ine and so tend to V-shapes. The rastered examp e is shown without a bonded id whi e the vectored channe s are shown after bonding to enc ose them. Maximum channe width and height dimensions of the enc osed vector channe s were obtained from image ana ysis ( mageJ) for a constant aser speed of ~ 0.3 ms - Fig. 8. The cross-sectiona area was determined using the mageJ tracing function and area versus aser power is observed to be inear. Area va ues obtained using a trapezoid or triangu ar profi e showed average errors of 2% (± 3%) and 6% (± 3%) respective y. The rastered snake channe cross-sectiona area was 7.35 x 0 -8 m 2 which is > 5x the argest vector area and measured vo umes ranged from ~ 8.5 µL to 73.5 µL.   Metering ratios were determined using K standard so utions (Hanna nstruments) with high ( 350 -550 µS/cm) and ow (84 -95 µS/cm) conductivity and D water ( -2 µS/cm). onductivity va ues were confirmed using a temperature-compensated Metrohm 7 2 probe meter. An vef ow AF -P 600 pressure generator was used to drive f uid through microf uidic devices. Two experimenta schematics are shown in Figs. 9 and 0 for snake channe s and Y meters respective y.
vef ow f ow meters with maximum rates of mL minand 5 mL minunder S software contro were used to measure f ow rates for known positive pressures and the hydrodynamic resistance R h obtained from the inverse s ope of f ow rate versus pressure. For snake ce s high conductivity (~ 500 µS/cm) and D water so utions were simu taneous y pumped using ce s of different engths positioned in each arm of the f uidic set up Fig. 9. The outputs were combined at an equa T-piece and the resu tant conductivity measured in a 0 mL ce containing a temperature compensated 7 2 Metrohm conductometer probe. The metering ratio is defined as the vo ume ratio of the two so utions (MR = V / V 2 ) where V represents the seawater equiva ent so ution and V 2 represents the high conductivity acid so ution. The fina tota vo ume is V T = V + V 2 . y c osing the V 2 channe the va ue of V can be determined. A ternative y by assuming the same re ationship between ion concentration and conductivity for a so utions then s T V T = s V + s 2 V 2 which can be rearranged to give  Tests were a so performed under negative pressure where the two f uids are simu taneous y drawn through the devices so representing the preferred configuration for an ocean dep oyed device for acid/seawater mixing. etoni eMySys syringe pumps were used with precision g assbodied Setonic syringes as in Fig. . Here syringe pu s f uid from both high and ow conductivity so utions through the Y meter test device and then dispenses 8 mL of the mixed so ution into the conductometer probe ce . A uminised Dak aPack L spoutbags specified as suitab e for gas ess iquids were used to ho d the f uids. The caps were fitted with Diba PTF 2-way va ves and si icone app ied to prevent f uid eakage. Syringe 2 in Fig. takes f uid from the ow refi bag and rep enishes the ow so ution bag to its initia eve immediate y after each mix samp e is taken. Refi of the high so ution bag is not necessary since the ow:high mix ratios is > 0: .

S mul o
The theoretica hydrodynamic resistance R H of a rectangu ar cross-section channe depends on width w and height h as in equation ( ). For circu ar cross-section of radius R it depends on R -4 . The high sensitivity of R H to dimension and shape means that routine methods of estimating R H for aser etched cross-sections are not avai ab e. Finite e ement mode ing of f uid f ow in channe s of various cross-sectiona shapes and dimensions was carried out using omso Mu tiphysics software (v5.3). Snake channe hydrodynamic resistance was determined by mode ing one comp ete straight portion of 30 mm ength and turn of diameter 2 mm then mu tip ying the resu t by the number of turns. hanne cross-sections were mode ed using a trapezoida approximation with varying base widths W or using a curved profi e obtained from a fit to the imaged cross sections. The rastered channe cross-section is shown in Fig. 2 with a base width of ~ 40 µm estimated from the f at region and a comparison between trapezoid and smoothed fitted curve profi es is shown in Fig. 3. The maximum channe height and width are 352 m and 298 m respective y. From simu ation R h increases as the base width of the trapezoid or fitted profi e is reduced towards a triangu ar cross-section with maximum of change of > 200% Fig. 4. The resistance obtained from the fitted profi e was up to 25% ower and ess sensitive to base width. Resistances from both profi es are equa when the curved profi e base width is ~40 m ess than that of the trapezoid Fig. 5.

4.
Di c io Graphs of f ow rate versus differentia pressure for each of the 4 snake channe s were found to be inear for a devices an examp e being shown in Fig. 6. The hydrodynamic resistance R H obtained from the inverse gradient is given in Fig. 7 and shows for constant cross-sectiona area R H proportiona to snake channe ength as expected. Simu ated va ues of R H using a trapezoida cross-section with a base width of 40 µm and depth of 298 µm as obtained from imaging indicate a 5% reduction compared to the measured va ues and with a fitted curve cross-section the simu ated va ue is ower by a further ~ 0%. One possib e reason for this discrepancy is the reduction of channe height and possib e channe deformation observed after PMMA bonding where we have previous y observed 5 µm -25 µm height oss after H 3 vapour assisted bonding. [42]. The simu ated impact of height oss on R H is shown in Fig. 8 for the trapezoida approximation where the 5% resistance difference is equiva ent to 23 m height reduction. n achieving narrower channe s for the high resistance acid ine we observe a change in profi e from trapezoida to a most triangu ar with dimensions given in    Metering ratios with snake channe s in both high and ow conductivity ines were determined from conductivity measurements and compared to those ca cu ated from R H measurements from the same devices. A 320 mm snake was inserted in the high conductivity ine (acid equiva ent) whi e the ow conductivity ine was formed from each of the shorter snake channe s in turn. The operationa positive pressure was 60 m ar. Tab e ists the measured meter ratios for each of the 3 versions with a maximum of > 4 for the shortest snake. Measured MR va ues and those predicted from R H agree to within ~4% i ustrating that the meter ratios can be predicted with reasonab e accuracy from the measured R H va ues. These ower MR va ues may be used where the avai ab e va ve techno ogy is not specified for ong term use with acid concentrations greater than ~ 0. M. Meter ratios were a so obtained from asymmetric Y-junction devices with one arm containing a short high resistance narrow channe obtained from vector mode aser etching with dimensions as given in Fig. 8. Measurements were taken using both positive ( vef ow) and negative ( etoni) pressure driven f ow and MR va ues from 7 to 6 were achieved in direct re ationship to the measured channe resistance Fig. 9. The sma difference between positive and negative pressure measurements in ike y due to the differences in externa tubing ength between both setups. Fina y the incorporation of an asymmetric Y-junction a ong with a sing e snake channe of various engths was used to increase the MR above 00 Fig. 20. This i ustrates the trade-off between area and MR. The area of the snake at the highest pattern reso ution wou d be > 600 mm 2 (6 x 0 -4 m 2 ) whi e that of the Y-meter is < 40 mm 2 . The use of narrow/sha ow vector mode channe s which have shown up to x35 increase in resistance per unit ength wou d reduce the area requirement by up to a factor of 0. However the integrity of ong and sha ow channe s has yet to be demonstrated within the constraints of a fu fabrication process as i ustrated here [42].  Meter ratio variabi ity was determined for each of the Ymeter devices given in Fig. 9. The rms precision va ues were found to be in the range 0. 5% -0.2 % whi e the so ution conductivity / conductometer precision was ~0. 0% Fig. 2 . These va ues are neg igib e with respect to impact on the overa T O 2 measurement precision Fig. 2. Fig. 2 . Re ative conductivity variation for 3 different Y meters with high conductivity so ution ( 4 3 µS cm -) in high resistance ine and ow conductivity so ution (87 µS cm -) in ow resistance ine. A so shown is conductivity variation of channe on y with direct dispense of K so ution (84 µS cm -) without Y-meter.

Co clus o
ontinuous and autonomous measurement of tota disso ved inorganic carbon (T O 2 ) in the oceans is critica for c imate change studies requiring miniaturised chemica ana ysis systems for ocean f oat-based dep oyment. A microf uidic conductivity-based approach offers the potentia for such miniaturisation and samp e pH reduction by acid injection to iberate disso ved O 2 is required which impacts on tota device pay oad. To address this we have fabricated aser etched microf uidic snake channe restrictors and asymmetric Y meters to adjust the metering ratio between samp e and acid reagent. Laser etching conditions were varied to create a range of channe dimensions down to ~75 m. hanne f ow versus pressure measurements were used to determine hydrodynamic resistances which were compared with finite e ement simu ations using a range of cross-section profi es and areas. Standard ( aser raster mode) channe s disp ayed specific R H va ues of 8 x 0 2 Pa.s m -3 whi e narrow/sha ower channe s ( aser vector mode) were 40 -300 x 0 2 Pa.s m -3 . Microf uidic metering circuits were constructed from variab e resistance snake channe s and dimensiona y symmetric or asymmetric Y-junctions. Samp e to acid vo ume ratios (meter ratio) up to 00: have been achieved with 300 m wide snake channe for engths > m. At the highest pattern reso ution this wou d require a footprint of > 600 mm 2 (6 x 0 -4 m 2 ). ircuits based so e y on asymmetric Y-junctions gave meter ratios up to 6: with a footprint cost of < 40 mm 2 and precision va ues of ~0.2%. Meter ratio variabi ity was neg igib e with respect to overa T O 2 ana ysis compared to the impact of temperature via reagent and samp e viscosity changes. Temperature dependence can be negated with the use of high MR ratios above 75. Further design and fabrication refinements wi be required to ensure the structura and dimensiona integrity of such sma channe s in future integration of metering units into fu T O 2 ana ysis microf uidic circuits.