Gold Nanoparticle Films As Sensitive and Reusable Elemental Mercury Sensors

Gold Nanoparticle Films As Sensitive and Reusable Elemental Mercury Sensors

Jay Z.James,?Donald Lucas,?,*and Catherine P.Koshland§

?Department of Mechanical Engineering,University of California at Berkeley,

?Environmental Energy Technologies Division,Lawrence Berkeley National Laboratory,

§School of Public Health,University of California at Berkeley,Berkeley,California94720,United States

*Supporting Information

concentrations from1to825μg Hg/m air3.Increasing the?ow velocity(and

system a viable method for direct ambient mercury vapor

heating to160°C,allows for repeatable measurements on the same

INTRODUCTION

Mercury,a neurotoxic global pollutant,demands global regulation.The long lifetime of mercury in the atmosphere (>1year)allows long-range transport limiting local emission controls from protecting all environments.Policy makers are working toward a worldwide e?ort similar to the sulfur dioxide or CFC regulations of the20th century.Anticipating a global policy,the European Commission began a?ve-year project in 2010called the Global Mercury Observation System(GMOS, www.gmos.eu)to create a coordinated global network adequate for improving models and making policy recommendations. The new system expands on the regional e?orts made in North America(i.e.,the Mercury Deposit Network and North American Airborne Mercury Experiment)and the independent observations made around the world.A preliminary assessment by GMOS points to gaps in emissions monitoring and in the spatial coverage of environmental observations,mostly in the southern https://www.sodocs.net/doc/fc7405302.html,ck of an inexpensive,stand-alone,low-power,low-maintenance sensor is a primary technical issue confronting the GMOS.

Current air monitors are amply sensitive to detect the global background(1.2ng/m3)but are costly and high maintenance.1 Preconcentration of trace mercury vapor samples is required for ambient measurements and depends on gold surfaces to e?ectively trap and quickly release mercury in detectable amounts.Concentrations as low as0.1ng/m3can be analyzed using a trap made of gold-coated sand that feeds a cold vapor atomic?uorescence or absorption spectroscope.2,3

Beyond providing selective adsorption for sample collection, gold performs as a transducer for mercury in various measurement techniques.Thundat et al.(1995)demonstrated the quantitative detection of adsorbed mercury with gold-coated microcantilevers.4Picogram resolution was obtained by monitoring changes in the resonant frequency of the cantilever bending.Resistivity changes in gold?lms are proportional to the adsorbed mercury mass,5an e?ect utilized by commercially available mercury vapor analyzers with～μg Hg/m air3resolution. Adsorbed mercury also induces changes in the optical properties of gold?lms,measurable with surface plasmon resonance spectroscopy.6In nanoparticle form,the optical property changes can be followed with simple and inexpensive visible light absorbance spectroscopy.Metal nanoparticles exhibit peaks in absorbance due to localized surface plasmon resonance(LSPR).The location of the peak depends on shape, size,composition,and local environment.Adsorption of mercury alters a gold nanoparticle’s complex dielectric function and causes a blue shift in the LSPR wavelength,seen in both

Received:February9,2012

Revised:June18,2012

Accepted:August7,2012

Published:August7,2012

colloidal solutions7?10and?lms.11,12Our recent work studying the LSPR response of individual gold nanorods achieved attogram(10?15g)resolution and found the sensitivity to be proportional to the nanoparticle’s surface-area-to-volume ratio.13Responding to the technical needs of an expanding mercury observation network,we are developing a reliable, sensitive,and inexpensive method for LSPR-based mercury detection.

■EXPERIMENTAL SECTION

Nanoparticle Film Preparation.Quartz,diced in9mm squares,acts as the transparent substrate for the nanoparticle ?lm.Before use,the quartz surfaces were cleaned in piranha solution for15min,(Caution!Piranha is a strong oxidizer and should not be stored in closed containers)rinsed in water(18.2 MΩ,Millipore)and ethanol,and dried in nitrogen.We used commercially available4-tert-butylthiophenol functionalized2?5nm gold nanoparticles(Alfa Aesar).Particle measurements (ImageJ)from the TEM images reveal the particles to have an average diameter of4.3nm with a standard deviation of2.5nm.

A histogram of the nanoparticle size distribution is available in the Supporting Information.The nanoparticles were suspended in chloroform and deposited,dropwise,onto the water surface held by a Te?on Langmuir?Blodgett trough(Nima).After30 min,the?lm was compressed,using the motorized Te?on barrier,to15mN/m surface pressure.The particle monolayer forms while?oating on the aqueous subphase,controlled by the uniaxial compression of the trough surface area.The substrate dipper then drew the submerged quartz chips and TEM grids (silicon nitride,Ted Pella)through the?oating nanoparticle layer?xing the particles to the substrate surfaces while the barrier holds the?lm at a uniform compression.

Film Characterization.TEM imaging(H-7650,Hitachi) and UV?vis absorption spectroscopy(HR4000,Ocean Optics) provided characterization of the particle?lms.A Lorentzian curve,?tted to the recorded spectra using Matlab,located the peak wavelength of the LSPR with a resolution of0.5nm. Sample Bag Method.Initial exposures to mercury vapor employed a Te?on sample bag(SKC)with a controlled dilution of saturated mercury vapor in clean air(Zero Air, AirGas).A peristaltic pump drew the sample from the bag over the sensor chip at a constant?ow of15cc/min.A quartz?ow cell(Starna Cells)held the sensor chip for in situ recording of the absorbance spectra.Each spectrum saved is the average of 100spectra with integration times of80ms.This technique was used for samples ranging from25to825μg Hg/m air3.The concentration of each dilution was measured using a conductometric mercury analyzer with an accuracy of±5% (Jerome J405).

Permeation Tube Method.For higher?ow rates and lower concentrations a chip was a?xed with a silicone adhesive to a1.25cm inner-diameter Pyrex tube such that the?lm faced normal to the axis of the tube.The collimating lenses and tube were held in a?xed position with the beam perpendicular to the sensor chip,ensuring observation of a consistent area of the chip during the absorbance measurements.

A permeation tube(VICI Metronics)in a steady?ow of air supplied a constant mercury concentration for the higher?ow rates.The emission of Hg from the permeation tube is constant for a given temperature with60ng/s emitted at room temperature(295K).In a stream of57L per minute(LPM) of air,the permeation tube system provides1.05μg Hg/m air3.

Regeneration.Heating tape,wrapped about the tube and connected to an autotransformer,was used to regenerate the sensor.A low?ow of mercury free air(6LPM)during heating purged the system.A low temperature(433K)for regeneration kept the nanoparticles from coalescing allowing reuse of the

?lm for further measurements.

■RESULTS AND DISCUSSION

Optimizing LSPR-based mercury sensing requires choosing the best material for selective adsorption and sensitive response. We used gold because it is a selective and stable mercury adsorbing material14,15and can be grown in a variety of shapes and sizes of nanoparticles.16We then determined the most sensitive and stable gold nanoparticle from available shapes and sizes.Morris et al.exposed?lms of gold nanospheres of varying size to saturated mercury vapor in room temperature air.12 They found that the shifts in LSPR at saturation were greater for smaller particles.Our previous work observing individual gold nanorods’spectral response toμg/m3concentrations of mercury in air found that the sensitivity was not dependent on size directly,but proportional to surface-area-to-volume ratio.13 We selected～5nm spheres because they have the largest surface-area-to-volume ratio while still having an observable peak in absorbance for an assembled?lm.12A schematic of the sensor is shown in Figure1.Spheres are the minimum surface-

area-to-volume ratio shape but they can be synthesized in much smaller sizes than the more complicated geometries.Being the minimum surface-area-to-volume geometry also serves as an advantage for the shape stability of spheroid particles.

The response of a5nm AuNP’s LSPR to amalgamation can be predicted with the use of existing models and previous experiments.The LSPR wavelength of bimetallic nanoparticles shifts proportionally to alloying mass fraction.17Because of di?erences in the complex dielectric,a5nm Hg particle LSPR wavelength would be273nm,240nm shorter than a AuNP of the same size.The Link et al.model predicts a shift of2.4nm for each percentage increase in the Hg mass fraction,which

in Figure1.Schematic of gold nanoparticle?lm during exposure to mercury vapor.

the case of the5nm sphere is equivalent to38atoms of Hg. The model agrees with experimental observations comparing UV?vis spectra with the measured mass fraction.7,10,13 Whereas we have observed the spectra of individual nanoparticles,both AuNP and amalgam particles,with dark ?eld spectroscopy,for a practical sensor we measure the spectral response of an array of particles using UV?vis absorbance spectroscopy.Assembly of such a particle?lm can be done in a variety of ways.We decided to use the Langmuir?Blodgett method because we can manufacture nanoparticle ?lms in a controlled,parallel,inexpensive fashion.18TEM images show the AuNP-?lm to be near close-packed with particles having an average diameter of4.3nm.A typical TEM images is shown in part a of Figure2.An average density of particles found in the TEM images from di?erent locations on the grid(using ImageJ’s particle counting routine)was used to approximate the total number of particle on the sensor chip. The result suggests that there are about2×1012particles(18μg)on each1cm2chip.The close proximity of the particles allows coupling between neighboring plasmons driving the resonance to longer wavelengths(isolated particles have an LSPR wavelength of～520nm).19,20The red shift in the LSPR is accompanied by a peak broadening shown in part b of Figure 2.All?lms tested originate from the same Langmuir?Blodgett batch and have an average LSPR wavelength of547.5nm.Films exposed to mercury vapor exhibit a blue shift in their LSPR wavelength that slows as the chip saturates as in Figure3. The LSPR shift as a function of time(Figure3)represents an integral of mercury concentration.Thus,a time derivative of the LSPR shift corresponds to a continuous measurement of concentration.Exposed to varying concentrations of Hg in air, the?lms generated using the Langmuir?Blodgett technique blue shift faster for higher https://www.sodocs.net/doc/fc7405302.html,ing a?ow rate of 15cc/min and mercury concentrations ranging from25to825μg Hg/m air3the initial LSPR shift rates(v LSPR)are proportional to the sample concentrations.We used a linear?t of the?rst2 nm of the peak’s blue shift to determine the initial v LSPR.The shift rates for5?lms exposed to a range of concentrations show a correlation in part a of Figure4.A linear regression of the mercury concentration dependence of v LSPR,?xed at the origin, predicts eachμg Hg/m air3increase in sample concentration will result in a0.023nm/h increase in v LSPR.

The sensitivity of the sensor depends on the interparticle distance within the?lm.Despite controlling for a constant surface pressure during the deposition(15mN/m),the prepared?lms have varying initial LSPR wavelengths(541.5?550nm).The longer LSPR wavelengths correspond to greater coupling between the nanoparticles in the?lm caused by smaller interparticle spacing.20Because the?lms were exposed to di?erent mercury concentrations,we use a normalized v LSPR

as a proxy for the sensitivity.The normalized v LSPR is the ratio of the v LSPR and the mercury concentration,and has units of nm/h/(μg/m3).Plotted in part b of Figure4,the shorter initial LSPR wavelengths show greater sensitivity with the normalized v LSPR decreasing67%in the range of initial LSPR tested.The decrease in sensitivity can be attributed to a combination of mass transfer and optical e?ects.In terms of the latter,the data suggest that plasmon coupling arising from closely spaced nanoparticles reduces the sensitivity of the?lm LSPR to adsorbed mercury mass.Given a probed area of?xed dimensions,the number of particles decreases with increasing interparticle spacing.Probing fewer particles reduces the mass sensitivity of the sensor.Increased interparticle spacing within the range tested here should increase the mass?ux of mercury per particle.Adsorption at neighboring particles depletes the local concentration retarding adsorptive?ux per particle.The results support using a lower surface pressure during deposition to generate more disperse?lms.Tao et al.(2007)previously demonstrated the role of surface pressure during the Langmuir?Blodgett deposition in determining the packing density of silver nanoparticles.21By reducing the surface pressure to8mN/m,we were able to reduce the packing density and create?lms with resonant absorbance at～520nm.

A representiative absorbance spectrum and TEM image of these less-dense?lms are available in the Supporting

Information.

Figure2.(a)TEM image of the Langmuir?Blodgett generated?lm showing the particle sizes and distribution in?lm.(b)A typical UV?vis absorption of the nanoparticle?lm on quartz.The?lm’s LSPR wavelength is550nm.

The v LSPR is also proportional to the rate of mercury adsorption,which is controlled by the di ?usive mass transfer of the trace mercury vapor.The mass transfer rate determines the time resolution of the sensor and can be controlled through engineering the exposure conditions.Increasing the sample

?ow rate reduces the collection times for low concentration measurements.For a ?at surface introduced into a uniform ?ow ?eld,?owing parallel to the surface,the mass transfer at the surface for a single dissolved species can be solved analytically if the surface concentration is known and the ?ow remains laminar.Elemental mercury vapor has a high a ?nity for gold,with an observed sticking coe ?cient of approximately one.This allows the assumption that the mercury vapor concentration at the gold boundary is zero.The solution predicts a mass transfer rate proportional to the square root of the bulk velocity,and linear with the mercury concentration.However,the square root dependence does not hold for turbulent ?ows.Experi-ments using the same geometry with turbulent ?ows show that the mass transfer rate remains directly proportional to concentration and proportional to the ?uid ’s Reynolds number to a power ranging between 0.8and 1.

The need for continuous measurements and remote operation make a single-use sensor chip limited in appeal.It is well-known that bulk gold releases mercury when heated to above 373K,and the surface can be regenerated to collect gold in a repeatable fashion.It is not clear that gold nanoparticles have the same property,especially for nonspherical shapes.We found that gentle heating of the amalgam nanoparticle ?lm evolves mercury as a vapor.The sensor response to mercury exposure following an hour at 433K was consistent,with no degradation observed for more than 30regenerations.This process is similar to the preconcentration of mercury samples using a gold trap,but at a lower temperature.The bulk gold ?lms used in the traps are more structurally stable and can survive the 1173K temperatures they are commonly heated to during the evolution step.Preservation of the nanoparticle ?lm morphology is key to a reproducible LSPR response;the melting point is size dependent and melting point depression in nanoparticles makes the nanoparticle ?lm vulnerable to lower temperatures than bulk gold.The depression is nonlinear with particle dimension,greatly impacting the melting temperature of our 4.3nm particles.22Heating to 513K for one hour causes irreversible changes to the ?lm wherein the particles coalesce forming larger particles (d =8nm).A TEM image of one such ?lm is shown in Figure 5.The larger diameter particles show a 50%reduction in sensitivity to mercury vapor,due in part to a reduction of the surface-area-to-volume ratio.

A single sensor chip was used in a series of exposures and regenerations to test the precision of the method and its agreement with ?ow rate trends of the mass transfer model.For six runs done with a ?ow rate and concentration of 20LPM and 3μg/m 3,the v LSPR was 1.1nm/hour on average with a standard deviation of 7%.Increasing the ?ow rate to 57LPM with a sample of 1μg Hg /m air 3the v LSPR was 0.42nm/hour.For ?ow rates below 40LPM,the ?ow in the Pyrex tube is laminar.Figure 6demonstrates good agreement between the mass transfer model and the experimental data by normalizing v LSPR to a single concentration (1ng Hg /m air 3,a typical ambient concentration 23).The experimental results follow a square root dependence on ?ow rate for ?ows up to 40LPM,but for 57LPM ?ow (which is expected to be turbulent (Re =3584))the v LSPR is three times faster than the laminar trend predicts.The time resolution of LSPR sensing of ambient mercury needs to be competitive with the existing methods,some of which take 24h of collection.The time resolution is limited by the rate of adsorption,which increases with Reynolds number.At the greatest ?ow rate tested,57LPM,an ambient mercury measurement (1ng Hg /m air 3)would take 410h to shift 1

nm.

Figure 3.Time trace of relative LSPR peak position during exposure to mercury vapor (3μg/m 3@20LPM).A linear ?t to the ?rst nanometer of shift is used to determine v LSPR for the

run.

Figure 4.(a)Calibration curve generated using nanoparticle ?lms exposed to a constant stream of 15cc/min mercury vapor (20?825μg/m 3)with an insert magnifying the results from lower concentration exposures.(b)The same experimental results as (a)but plotted to show the dependence of the v LSPR on the initial LSPR wavelength.Normalized v LSPR is ratio of the measured v LSPR to the mercury vapor concentration.

By accelerating the ?ow rate or implementing an impinging ?ow,the time resolution can be reduced dramatically.

Accuracy of the sensor can be improved by controlling for the confounding factors.Observation of LSPR temperature dependence during the heating and cooling steps of regeneration prompted the use of a thermocouple to monitor the sensor temperature.A linear regression of the LSPR versus temperature data from the hour before exposure allows normalization of the peak position;the LSPR peak temperature dependence was 1.7nm/K.Additional confounding e ?ects appear as a gradual red shifting (0.02nm/h)of the LSPR for mercury free sample air.This is likely due to other adsorbates that increase the index of refraction surrounding the AuNPs causing the shift of the resonance to longer wavelengths.No e ?orts were made to correct for the red shifts,as they are slower than the standard deviation of the sensor response to the tested mercury concentrations (7%).Confounding e ?ects will increase with respect to the mercury signal at lower

concentrations but have been mitigated through the use of protective monolayer ?lms on similar gold ?lm based sensors.24With the aim of developing a highly sensitive and inexpensive mercury vapor sensor we employed a ?lm of ～5nm AuNPs in an LSPR-based chemical sensor.AuNP ?lms act as both receptors and transducers by selectively adsorbing mercury vapor and changing their visible absorption spectra.Response to concentrations between 1to 825μg Hg /m air 3con ?rms that the LSPR method is equivalently sensitive to current commercial methods.Adsorption of 15Hg atoms per nanoparticle causes a 1nm shift in LSPR wavelength,twice the error of our Lorentzian peak measurement (0.5nm).Multiplying the number of Hg atoms per particle by the number of particles in the optical path (2×1012particles/cm 2in an area of ～1mm 2)yields 100picograms of mercury causing the observed changes.Simple heating of the ?lm causes a reversible desorption of mercury.The cost of such a monitor comprised of a AuNP ?lm,heater,pump,lamp,and appropriate spectrometer is considerably less than the methods used now.

■ASSOCIATED CONTENT

*

Supporting Information TEM images,and UV ?vis absorbance data.This material is available free of charge via the Internet at https://www.sodocs.net/doc/fc7405302.html,.

■AUTHOR INFORMATION

Corresponding Author *E-mail:d_lucas@https://www.sodocs.net/doc/fc7405302.html,.

Notes

The authors declare no competing ?nancial interest.

■ACKNOWLEDGMENTS

This project was supported by Award Number P42ES004705from NIEHS and the Wood-Calvert Chair.The content is solely the responsibility of the authors and does not necessarily represent the o ?cial views of NIEHS or NIH.

■

REFERENCES

(1)French,N.B.;Priebe,S.J.;Haas,W.J.,Jr.Product reviews:State-of-the-art mercury CEMs.Anal.Chem.1999,71,470A ?475A.

(2)Hatch,W.R.;Ott,W.L.Determination of submicrogram quantities of mercury by atomic absorption spectrophotometry.Anal.Chem.1968,40,2085?2087.

(3)Muscat,V.I.;Vickers,T.J.;Andren,A.Simple and versatile atomic fluorescence system for determination of nanogram quantities of mercury.Anal.Chem.1972,44,218?221.

(4)Thundat,T.;Wachter,E.;Sharp,S.;Warmack,R.Detection of mercury-vapor using resonating microcantilevers.Appl.Phys.Lett.1995,66,1695?1697.

(5)McNerney,J.J.;Buseck,P.R.;Hanson,R.C.Mercury detection by means of thin gold films.Science 1972,178,611?612.

(6)Morris,T.;Szulczewski,G.Evaluating the role of coinage metal films in the detection of mercury vapor by surface plasmon resonance https://www.sodocs.net/doc/fc7405302.html,ngmuir 2002,18,5823?5829.

(7)Morris,T.;Copeland,H.;McLinden,E.;Wilson,S.;Szulczewski,G.The effects of mercury adsorption on the optical response of size-selected gold and silver https://www.sodocs.net/doc/fc7405302.html,ngmuir 2002,18,7261?7264.(8)Scallan,K.;Lucas,D.;Koshland,C.A novel UV-vis spectroscopic method to detect elemental mercury using gold nanoparticles.Epidemiology 2006,17.

(9)Rex,M.;Hernandez,F.E.;Campiglia,A.D.Pushing the limits of mercury sensors with gold nanorods.Anal.Chem.2006,78,445?451.(10)Mertens,S.F.L.;Gara,M.;Sologubenko,A.S.;Mayer,J.;Szidat,S.;Kra m er,K.W.;Jacob,T.;Schiffrin,D.J.;Wandlowski,T.Gold ?mercury nanoalloys:Au@Hg nanoalloy formation

through

Figure 5.TEM image of nanoparticle ?lm showing irreversible changes following heating to 513K for 1

h.

Figure 6.Normalization of all tests to a single concentration (1ng/m 3)to demonstrate the e ?ect of the sample ?ow rate.The experimental data (open circles)overlaid on the mass transfer models,laminar (solid line),and turbulent (dashed line).

direct amalgamation:Structural,spectroscopic,and computational evidence for slow nanoscale diffusion(Adv.Funct.Mater.17/2011). Adv.Funct.Mater.2011,21,3202.

(11)Wang,C.;Ma,L.;Hossain,M.;Wang,H.;Zou,S.;Hickman,J. J.;Su,M.Direct visualization of molecular scale chemical adsorptions on solids using plasmonic nanoparticle arrays.Sens.Actuators,B2010, 150,667?672.

(12)Morris,T.;Kloepper,K.;Wilson,S.;Szulczewski,G.A spectroscopic study of mercury vapor adsorption on gold nanoparticle films.J.Colloid Interface Sci.2002,254,49?55.

(13)James,J.Z.Mercury sensing with optically responsive gold nanoparticles Ph.D.Dissertation,UC Berkeley,Berkeley,CA,2012. (14)Braman,R.S.;Johnson,D.L.Selective absorption tubes and emission technique for determination of ambient forms of mercury in air.Environ.Sci.Technol.1974,8,996?1003.

(15)Levlin,M.;Ika v alko,E.;Laitinen,T.Adsorption of mercury on gold and silver surfaces.Fresenius’Journal of Analytical Chemistry1999, 365,577?586.

(16)Grzelczak,M.;Perez-Juste,J.;Mulvaney,P.;Liz-Marzan,L.M. Shape control in gold nanoparticle synthesis.Chem.Soc.Rev.2008,37, 1783?1791.

(17)Link,S.;Wang,Z.L.;El-Sayed,M.A.Alloy formation of gold?silver nanoparticles and the dependence of the plasmon absorption on their composition.J.Phys.Chem.B1999,103,3529?3533.

(18)Tao, A.R.;Huang,J.;Yang,https://www.sodocs.net/doc/fc7405302.html,ngmuir?Blodgettry of nanocrystals and nanowires.Acc.Chem.Res.2008,41,1662?1673.

(19)Jain,P.K.;Huang,W.;El-Sayed,M.A.On the universal scaling behavior of the distance decay of plasmon coupling in metal nanoparticle pairs:A plasmon ruler equation.Nano Lett.2007,7, 2080?2088.

(20)Sendroiu,I.E.;Mertens,S.F.L.;Schiffrin,D.J.Plasmon interactions between gold nanoparticles in aqueous solution with controlled spatial separation.Phys.Chem.Chem.Phys.2006,8,1430?1436.

(21)Tao, A.;Sinsermsuksakul,P.;Yang,P.Tunable plasmonic lattices of silver nanocrystals.Nat Nano2007,2,435?440.

(22)Buffat,P.;Borel,J.-P.Size effect on the melting temperature of gold particles.Phys.Rev.A1976,13,2287?2298.

(23)Slemr,F.;Brunke,E.-G.;Ebinghaus,R.;Kuss,J.Worldwide trend of atmospheric mercury since1995.Atmos.Chem.Phys.2011, 11,4779?4787.

(24)Mirsky,V.M.;Vasjari,M.;Novotny,I.;Rehacek,V.;Tvarozek, V.;Wolfbeis,O.S.Self-assembled monolayers as selective filters for chemical sensors.Nanotechnology2002,13,175.