A combined nanoplasmonic and electrodeless quartz crystal microbalance setup

REVIEW OF SCIENTIFIC INSTRUMENTS 80, 125105 共2009兲

A combined nanoplasmonic and electrodeless quartz crystal microbalance setup Elin M. Larsson,1,a兲 Malin E. M. Edvardsson,2 Christoph Langhammer,1 Igor Zorić,1 and Bengt Kasemo1 1

Department of Applied Physics, Chemical Physics Group, Chalmers University of Technology, SE-41296 Gothenburg, Sweden 2 Department of Applied Physics, Biological Physics Group, Chalmers University of Technology, SE-41296 Gothenburg, Sweden

共Received 1 June 2009; accepted 27 October 2009; published online 7 December 2009兲 We have developed an instrument combining localized surface plasmon resonance 共LSPR兲 sensing with electrodeless quartz crystal microbalance with dissipation monitoring 共QCM-D兲. The two techniques can be run simultaneously, on the same sensor surface, and with the same time resolution and sensitivity as for the individual techniques. The electrodeless QCM eliminates the need to fabricate electrodes on the quartz crystal and gives a large flexibility in choosing the surface structure and coating for both QCM-D and LSPR. The performance is demonstrated for liquid phase measurements of lipid bilayer formation and biorecognition events, and for gas phase measurements of hydrogen uptake/release by palladium nanoparticles. Advantages of using the combined equipment for biomolecular adsorption studies include synchronized information about structural transformations and extraction of molecular 共dry兲 mass and degree of hydration of the adlayer, which cannot be obtained with the individual techniques. In hydrogen storage studies the combined equipment, allows for synchronized measurements of uptake/release kinetics and quantification of stored hydrogen amounts in nanoparticles and films at practically interesting hydrogen pressures and temperatures. © 2009 American Institute of Physics. 关doi:10.1063/1.3265321兴

I. INTRODUCTION

Progress in surface science has to a large extent been driven by the development of new and improved experimental probes to study surfaces and surface processes. With continuously growing complexity of the surface systems under study, there is increasing need for more diversified information. For comprehensive understanding of a studied system the information obtained with a single technique is often insufficient and the introduction of additional, complementary technique共s兲 is a great advantage or a necessity. We report here on the development of a novel experimental setup that combines two surface sensitive techniques and allows real-time, label-free detection of molecular adsorption/absorption/desorption events: electrodeless quartz crystal microbalance (QCM) and optical localized surface plasmon resonances (LSPR) sensing. Optical measurements are made with a standard spectrometer by recording light transmitted through the sample 共simultaneously the QCM sensor兲 and measurement chamber. The performance of the setup and some of the advantages with the combination technique are demonstrated for both gas and liquid phase measurements. QCM1,2 and the special version QCM-D 共quartz crystal microbalance with dissipation monitoring兲3,4 are wellestablished techniques to study surface processes. Initially, QCM was used for measurements in ultrahigh vacuum and the gas phase, but in the past decades it has become a coma兲

Electronic mail: [email protected].

0034-6748/2009/80共12兲/125105/10/$25.00

mon technique for liquid phase measurement of, e.g., biomolecular interactions at surfaces and biointerfaces and5,6 of polymer and polyelectrolyte properties.7,8 The sensing principle is the detection of changes in the electromechanical characteristics of an oscillating piezoelectric quartz crystal sensor upon changes of the surface/interfacial properties at one or both of the crystal surfaces. The detected mass of, e.g., adsorbed biomolecular layers measured with QCM is not the “dry mass” of the biomolecules but also includes the mass of the solvent that is dynamically coupled to the interfacial film. Therefore, the latter is sometimes referred to as the “wet mass” or the “acoustic mass.”9 The oscillation of the crystal is driven by an ac voltage, which is usually applied between two electrodes evaporated on each face of the crystal. Optical techniques used for real-time, label-free detection of molecular adsorption/desorption events at surfaces include SPR spectroscopy,10 ellipsometry,11 reflectometry,12 reflectometric interferometric spectroscopy,13 and optical waveguide techniques.14 Although the exact implementation differs between these techniques, they are all sensitive to the same two parameters, namely the effective refractive index and thickness of an adsorbed layer. Since solvent coupled to the adsorbed layer has the same refractive index as the surrounding medium it is not measured using optical techniques. The measured mass is, therefore, sometimes referred to as the “optical” or “dry” mass. Optical and electromechanical 共QCM兲 measurements complement each other. For example, optical measurements

80, 125105-1

© 2009 American Institute of Physics

125105-2

Rev. Sci. Instrum. 80, 125105 共2009兲

Larsson et al.

in general give the accurate mass of the adsorbed molecules while electromechanical measurements give the mass of both the molecules and the coupled solvent, which is directly related to the structural properties of the adsorbed film. Therefore, combining QCM with an optical technique makes it possible to resolve the mass of both the adsorbates and the coupled solvent, as well as to gain additional information about structural changes in the adsorbed layers. Additional information about the adlayer, e.g., if it is significantly hydrated, or wet, is obtained with the QCM-D technique, where D stands for dissipation monitoring, or damping of the oscillation. Effectively D is related to the viscoelastic properties of the adlayer, which in turn depends on the degree of hydration.15 Although information from QCM共-D兲 and optical measurements can be obtained in separate experimental systems, it is more desirable, in many cases, to combine the techniques into one single instrument capable of analyzing the same surface concurrently. A combination instrument eliminates artifacts that may arise from nonidentical conditions, e.g., differences in sample preparation, handling and measurement conditions. Direct comparison of data is, therefore, simplified. This is of particular importance for surface based techniques where the sample preparation, handling, and environment are critical. An additional advantage of using combination instruments is that the time for collecting the necessary data is reduced. Optical techniques that have previously been combined with simultaneous QCM共-D兲 measurements, in the same instrument, include SPR,7,16,17 ellipsometry,18,19 reflectometry,20 and, in one case, LSPR sensing in nanoscopic holes in a thin 共30 nm兲 metal film.21,22 Except for LSPR sensing, all the techniques mentioned above employ optical reflection measurements. The latter usually requires complex optical equipment and modifications of the quartz crystal surface. In contrast, optical transmission measurements can be made using a standard spectrometer and there are no requirements on surface modifications. Alignment of the light beam is noncritical in contrast to, e.g., SPR, ellipsometry, and reflectometry. However, the combination of QCM with optical transmission measurements requires that the electrodes driving the oscillation of the quartz crystal are transparent or have a transparent region. In a previously reported QCM-D/LSPR setup the bottom electrode was made transparent by a hole in the center of the deposited electrode, while another deposited gold film, with plasmonic nanoholes, served as the top electrode.21,22 To obtain larger freedom for the optical transmission measurement, it would be desirable to eliminate the necessity of a simultaneously conductive and optically active film. Both the top and bottom electrodes would then need to have a hole 共or another optically transparent region, e.g., a grid兲, or be made of an optically transparent but conductive, material-like indium tin oxide. The latter is, however, complicated from a preparation point of view, and may also impede the function of the QCM crystal, e.g., due to the relatively low conductivity. The simplest and most attractive alternative, reported here, is to employ the so called electrodeless QCM, where

the electrodes driving the QCM sensor crystal are physically detached from the piezoelectric sensor crystal, and are placed a short distance away from the crystal surface, so that there is a gas or liquid filled gap between the sensor surface and the electrodes. The electrodes then become an integrated part of the experimental chamber, rather than of the sensor, and only need to be fabricated once. In addition to being attractive from a cost perspective, electrodeless QCM offers other advantages, since it allows for the sensor surface to be more freely designed with respect to the requirements of both the complementary technique and of the surface processes to be investigated. For example, conducting and nonconducting surfaces can both be used. It also eliminates any interference that may arise from chemical interaction of the electrode material with, for example, metal nanostructures 共e.g., alloying兲. II. THEORETICAL BACKGROUND A. QCM-D

The QCM is an electromechanical sensor used to detect small mass changes at surfaces both in the gas and liquid phase. An ac field is applied over a piezoelectric quartz sensor crystal in order to excite a mechanical shear oscillation in the crystal, at the fundamental or overtone frequencies, f z. An AT cut crystal, which is the most commonly used crystal cut, oscillates in what is called a thickness shear mode, i.e., with an oscillatory shear motion, which is parallel to the surface plane. The deposition—or loss—of small amounts of material on the crystal surface causes a decrease—or increase—in the resonance frequency, ⌬f z, of the quartz crystal and can, under certain conditions be related to the accumulation of mass, ⌬⌫QCM via the well-known equation formulated by Sauerbrey in 19592 ⌬⌫QCM = t · ␳ = −

C · ⌬f z , z

共1兲

where ␳ and t are the effective density and thickness of the adsorbed film, respectively. C is the mass sensitivity constant 关C = −17.7 ng/ 共cm2 ⫻ Hz兲兴 for a 5 MHz fundamental mode AT cut quartz crystal兲 and z 共=1 , 3. . .兲 is the overtone number. The Sauerbrey equation is strictly valid when the material adsorbed on the surface follows the shear oscillatory motion of the crystal rigidly as a “dead mass” and does not induce large energy losses 共dissipation兲 during the motion of the crystal. Determining the mass of strongly damped systems, e.g., viscoelastic layers composed of highly hydrated polymers or biomolecular structures 共e.g., gels or gel-like structures兲, which do not follow the shear motion rigidly but are deformed during the motion, requires combined frequency and energy dissipation measurements, D, as well as modeling of the coupling between these variables.15 The ac field, driving the oscillation of the quartz crystal, is usually applied via two gold electrodes evaporated on each side of the quartz crystal. However, so-called electrodeless QCM, where one or both of the electrodes have been lifted off the surface, has also been realized. Electrodeless QCM

125105-3

Larsson et al.

Rev. Sci. Instrum. 80, 125105 共2009兲

has been found to function as a mass sensor in a similar way as the electrode coated version in both the gas and liquid phases.23,24 B. Nanoplasmonic sensing

The optical properties of metals change drastically when their size is decreased below the wavelength of light. One particularly interesting property of metal nanoparticles, for e.g., sensing applications, is that an external electromagnetic field can cause the conduction electrons to oscillate collectively with respect to the atomic nuclei. The coupling between the electromagnetic field and the electrons is strongest at the LSPR wavelength. As a consequence of the LSPR excitation there is a strongly enhanced electromagnetic field generated around the nanoparticle, which acts as a probe of the dielectric environment near the particle 共typically within a few tens of nanometers兲. The spectral sensitivity of the LSPR to changes in the surrounding dielectric opens up for the possibility to use nanoparticles as transducers in sensors where they convert small refractive index changes in the particle nanoenvironment into spectral peak shifts. Although several application areas for refractive index-based nanoplasmonic sensors have been suggested, including metal oxidation,25,26 catalysis,27 and hydrogen storage,28 biosensing is the most widely investigated.29 Biomolecules in general have a larger refractive index than the solution they are in. In LSPR biosensors refractive index changes may, therefore, be caused by biomolecular adsorption or desorption events close to metal nanoparticles. LSPR-based biosensors have, for example, been applied to detect biomarkers for Alzheimer disease30 and oral cancer cells.31 Ultrasensitive quantification of proteins down toward single molecules has been demonstrated.29 In contrast to LSPR-based biosensing, LSPR-based hydrogen sensing,28 for the characterization of nanosized hydrogen storage systems, does not rely on detecting a change of the refractive index in the nanoenvironment of the sensor particle. Instead changes in the dielectric properties of the particle, induced by hydrogen absorption or desorption, are detected by reading out the LSPR response. The change in the electronic properties originates from the electronic interaction and charge exchange between the hydrogen atoms and the host lattice. The latter, in turn, causes a change in the atomic positions to minimize the energy in the system.32 A small contribution 共few percent of the total signal for the case of Pd兲 to the observed spectral response of the LSPR can also be related to a structural change of the nanoparticle upon hydrogenation. The structural change has its origin in volume expansion 共ca. 6%–7% for Pd-PdHx兲 to accommodate “space” for the hydrogen atoms and/or a change of the lattice structure during the hydride formation.32 III. INSTRUMENTATION

A measurement chamber, built in-house, that accommodates 1 in. diameter QCM crystals 共Lap-tech Inc., Canada兲 was used. The chamber is made of polymethyl methacrylate 共PMMA兲 and has one built-in metal 共brass/stainless steel兲 electrode on each side of the QCM crystal’s position,

FIG. 1. 共Color online兲 Schematic illustration 共not to scale兲 of the electrode configuration and the measurement setup for the combined electrodeless QCM/LSPR chamber.

hereafter referred to as the top and bottom electrodes, respectively. Figure 1 is a schematic illustration of the electrode configuration and measurement setup. The top and bottom electrodes are shaped as disks with a hole in the middle. They are mounted parallel to the QCM crystal at a distance of 0.3 mm. The outer diameters of the top and bottom electrodes are 21 and 16 mm, respectively, and the diameter of their holes is 5 mm. Optical transmission into the gas/liquid cell is facilitated by making the PMMA wall thin 共⬃3.5 mm兲 above the hole in the top electrode. The hole in the bottom electrode is uncovered since the area between the bottom electrode and the quartz crystal does not serve as a liquid or gas compartment. Gas/liquid inlet and outlet occurs through two holes 共⬃2 mm in diameter兲 in the top electrode. The inlet and outlet are located on opposite sides of the chamber, ⬃1 mm from the edge of the electrode, so that the gas/liquid flows over the crystal when moving from the inlet to the outlet. For liquid measurements, the outlet is connected to a peristaltic pump 共Reglo Digital MS 2/12, Ismatec, USA兲. To reduce disturbances arising from pressure changes when the pump is turned on and off to change solution, an extra liquid outlet is installed. A tube is connected to the extra outlet so that the liquid level in the tube can rise/sink to level out pressure changes inside the chamber. QCM measurements were made using a standard D300 共Q-Sense AB, Sweden兲 driving unit. A fiber-coupled spectrometer 共Avantes, the Netherlands兲 was used for optical transmission measurements. The maximum time resolution using the current equipment is ⬃10 ms for the optical measurement and ⬃70 ms for the QCM measurement. Plasmonic metal nanodisks and truncated nanocones were fabricated onto the blank QCM crystals using holemask colloidal lithography.33 This method yields a partly random distribution of nanostructures 共see Fig. 2兲 on the substrate, which, combined with a large particle-particle separation, eliminates both far and near field coupling between the particles so that the measured optical signal reflects the optical 共LSPR兲 response of a single particle.34 共The resulting

125105-4

Larsson et al.

Rev. Sci. Instrum. 80, 125105 共2009兲

IV. MATERIALS AND METHODS A. Liquid phase measurements

FIG. 2. 共Color online兲 共a兲 Schematic illustration of the sensing structure used for liquid phase measurements. 共b兲 SEM image of the quartz crystal sensor in 共a兲. 共c兲 Schematic illustration of the truncated palladium nanocone structures used to study hydrogen storage and release in the gas phase. 共d兲 SEM images of the palladium structures in 共c兲.

structures can be, e. g., disks or cones depending on the details of the fabrication.33 Both type of structures were used in the work presented below. The slight difference in nanostructure shape is not important in the present context.兲 The QCM crystal used for liquid phase measurement 关Figs. 2共a兲 and 2共b兲兴 was decorated with gold nanodisks that were 25 nm high and had an average diameter of ⬃76 nm. On top of the gold nanodisks a 3 nm chromium layer was deposited. The latter served as an adhesion layer for the SiO2 layer deposited in a subsequent step. Before SiO2 deposition, the nanodisk-covered crystal was heated for 2 h at 250 ° C in order to release any stress/strain, in the crystal, arising from the fabrication. For reliable supported lipid bilayer 共SLB兲 formation, a 20-nm-thick silica layer was, subsequently, sputtered 共reactive ion sputtering兲 onto the nanodisk-covered crystal, since this coating is the commonly used substrate for SLB formation.35 Figure 2共a兲 is a schematic illustration of the sensor surface after completed nanofabrication. The arrangement of the nanodisks on the surface is shown in the scanning electron microscopy 共SEM兲 image in Fig. 2共b兲. Gas phase measurements of hydrogen storage and release were made on a QCM crystal with truncated palladium nanocones. The height of the nanocones was 100 nm and the average top and bottom diameters 270 and 220 nm, respectively 关Fig. 2共c兲兴. The truncated cone shape of the structures is a direct consequence of the evaporation of a relatively thick metal film 共100 nm兲 through a hole mask with comparatively small holes 共250 nm diameter兲. For thinner films 共lower structures兲 and/or larger holes 共larger diameter of the structures兲 disks are obtained.33 After nanostructure fabrication the crystal was heated for 2 h at 220 ° C to stabilize the structure and relieve any remaining stress/strain, in the crystal, from the nanofabrication 共i.e., metal evaporation兲. A SEM image of the QCM crystal with truncated palladium cones is shown in Fig. 2共d兲. The inset is a SEM image of a single truncated palladium nanocone, showing clearly the polycrystalline nanostructure with an average grain size of ⬃25 nm.

1- Palmitoyl -2- oleoyl- sn - glycero -3- phosphocholine 共POPC兲 and 1,2-dipalmitoyl-sn-glycero-3-phosphoethanol amine—N—共cap biotinyl兲 共DPPE-biotin兲 共Avanti Polar Lipids, USA兲 were used to prepare POPC/DPPE-biotin 关95:5 共w/w兲兴 vesicles in Tris buffer 共10 mM Tris, 100 mM NaCl, pH 8兲, by extrusion. The lipid material was dissolved in chloroform 共stock solutions were stored at −20 ° C兲 and the solvent was evaporated by a flow of nitrogen 共⬎1 h兲 such that a thin lipid film formed on the wall of a glass flask. The lipid film was redissolved in buffer and extruded through polycarbonate membranes 共Whatman, USA兲, 25 times through membranes having a pore diameter of 0.1 ␮m, and additionally 15 times through membranes of pore diameter 0.03 ␮m. All vesicle solutions were stored at 4 ° C under nitrogen until use. Measurements were carried out at room temperature in Tris buffer 共10 mM Tris, 100 mM NaCl, pH 8兲. After mounting of the sensor crystal, the chamber was filled with Tris buffer at a flow rate of 50 ␮l / min. The flow was then maintained at the same rate throughout the measurement. When a stable baseline was obtained, biotinylated vesicles were added at a concentration of 0.5 mg/ml. After the completion of the bilayer the chamber was rinsed with buffer in order to remove excess vesicles. Streptavidin 共Sigma, Streptomyces avidinii, recombinant, expressed in Escherichia coli, freezedried, dissolved in water 共MilliPore, France, 1 mg/ml兲 and stored frozen in aliquots at −20 ° C until use兲 was added at a concentration of 10 ␮g / ml. After surface binding up to saturation had been reached, the chamber was rinsed with buffer. Next, biotinylated vesicles were added at a concentration of 0.5 mg/ml. Finally, the chamber was rinsed with buffer. For the QCM-D data presented here, z = 5. B. Gas phase measurements

Measurements were carried out at room temperature 共22 ° C兲 at atmospheric pressure using a gas flow rate of 60 ml/min. The gas flowing over the QCM crystal was alternated between 100% Ar and 4% H2 in Ar. For the QCM data presented here, z = 1. V. EXPERIMENTAL RESULTS AND DISCUSSION

To demonstrate the performance and capacity of the electrodeless QCM/LSPR combination, we performed measurements both in a conductive liquid 共buffer兲 and in gas phase. A. Liquid phase measurements

In the liquid phase the formation of a supported biotinylated lipid bilayer by deposition of vesicles on a silica surface was monitored. The lipid bilayer formation served as a well-defined reference system that has been described in detail previously35,36 and has been used to demonstrate the performance of QCM in combination with optical techniques such as SPR,37 reflectometry,9,20 and LSPR in nanoholes.22 To further demonstrate the capacity of the LSPR/QCM com-

125105-5

Larsson et al.

FIG. 3. 共Color online兲 共a兲 Schematic illustration of 共I兲 the formation of a biotinylated bilayer, 共II兲 streptavidin binding to the functionalized bilayer 共I兲, and 共III兲 binding of biotinylated vesicles to streptavidin. 共b兲 QCM-D data for the experiment in 共a兲. Vesicles enter the chamber and start adsorbing at t = 12 min. Streptavidin binding starts at t = 60 min and the binding of biotinylated vesicles starts at t = 126.5 min. Data from the fifth overtone 共z = 5兲 are shown. c兲 Frequency shift measured with QCM 关same data as in 共b兲兴 and LSPR peak shift during the experiment in 共a兲 and 共b兲. The inset is a magnification of the optical response during lipid bilayer formation. The solid lines are guides for the eye.

bination setup, molecular recognition and binding was shown by the specific binding of streptavidin to the biotinylated lipid bilayer. Subsequently, intact biotinylated vesicles were bound to the streptavidin. We are thus showing three subsequent steps: 共I兲 biotinylated SLB formation, 共II兲 streptavidin binding to 共I兲, and 共III兲 binding of biotinylated vesicles to 共II兲 as shown in Fig. 3共a兲. The latter steps have previously been used to investigate the performance of a combined reflectometry and QCM-D setup9 and, therefore, serve as a good reference system for the present LSPR sensing measurements. The result from a typical experiment is shown in Figs. 3共b兲 and 3共c兲. B. Lipid bilayer formation „step I…

The QCM-D response 共Fig. 3共b兲兲 upon exposure of the silica surface to biotinylated vesicles 共step I兲 shows the characteristic multiphase behavior seen previously.38 The initial adsorption of intact vesicles leads to a highly viscous layer with a high dissipation 共Ia兲. The lipid bilayer formed when

Rev. Sci. Instrum. 80, 125105 共2009兲

the vesicles rupture, and their fragments fuse, is a more rigid structure and has a very low dissipation compared to the adsorbed intact vesicles 共Ib兲. When the vesicles rupture, water that was previously trapped in the vesicles is released and the frequency increases. A final frequency shift of ⬃29 Hz and a dissipation of ⬃0.45⫻ 10−6 at saturation confirm the formation of a continuous bilayer that covers the nanofabricated support with only small defects such as trapped vesicles.38 Using the Sauerbrey equation, the mass of the bilayer was found to be 513 ng/ cm2 and its thickness 5.1 nm 共␳ = 1.004 g / cm3兲.39 This is slightly higher than previous results for a POPC bilayer,20,40,41 but is in agreement with the fact that there are defects such as trapped vesicles in the bilayer. Although the QCM response is an independent verification of the successful formation of a SLB, the optical response 关Fig. 3共c兲兴 provides information that allows the lipid bilayer formation to be studied in more detail. At t = 12 min, as intact vesicles bind to the silica, the optical signal redshifts almost monotonically due to the increased refractive index in the close vicinity of the gold nanodisks. Approximately 150 s later, at t = 14.5 min, there is a significant temporal increase in the rate by which the optical signal increases resulting in a “kink” in the curve 关in the inset in Fig. 3共c兲, two straight lines are inserted to guide the eye to the point where there rate of increase of the optical signal accelerates 共i.e., the “kink”兲兴. This optical signature of bilayer formation is attributed to the rupture process of adsorbed vesicles, which leads to a change in the average lipid distribution, with net lipid mass movement toward the surface, where the sensitivity of the plasmonic nanostructures is higher. Upon rupture of the vesicles, the lipids thus move into a region where the optical sensitivity is higher causing the plasmon peak to redshift rapidly.22 Another event, possibly contributing to the observed kink, is that the optical response may be sensitive to the orientation of the lipid molecules. Under such circumstances, vesicle rupture, leading to alignment of the lipids in a direction perpendicular to the substrate, may cause a rapid change in the optical signal. Molecular orientational sensitivity has been observed previously.40 At t = 20 min there is an abrupt change of the slope of the curve whereafter it levels off. This signifies that the adsorption of lipid material to the surface has stopped. Note that the optical response stabilizes somewhat earlier than the QCM responses, i.e., the net transport of lipid material to the surface stops before the bilayer formation process has completely terminated.37,42 1. Decay length of the LSPR field

With knowledge about the bilayer thickness, which was calculated from the frequency shift in the QCM measurement, and assuming an exponential decay of the LSPR field intensity it is possible to calculate the decay length of the enhanced electromagnetic field associated with the gold particle plasmon resonance. The effective refractive index sensed by the plasmonic particles, assuming an exponentially decaying field, is given by

125105-6

neff =

Rev. Sci. Instrum. 80, 125105 共2009兲

Larsson et al.

1 L

冕

y=⬁

共2兲

n共y兲e−y/Ldy,

y=0

where L is the decay length of the LSPR field intensity, n共y兲 the refractive index, and y the distance from the surface. For the situation with a SLB with the thickness t and the refractive index nSLB the effective refractive index 共neff兲 becomes neff =

nSLB L

冕

y=t

e共−y/L兲dy +

y=0

= nSLB关1 − e

共−t/L兲

nbuffer L

兴 + nbuffere

冕

y=⬁

e共−y/L兲dy

y=t

共−t/L兲

,

共3兲

where nbuffer is the refractive index of the buffer solution. The plasmon peak shift ⌬␭ induced by a change in the refractive index from nbuffer to neff is given by ⌬␭ = S共neff − nbuffer兲 = S关1 − e共−t/L兲兴共nSLB − nbuffer兲,

共4兲

where S is the bulk sensitivity of the sensor. The second step in Eq. 共4兲 was obtained by inserting the expression in Eq. 共3兲 for neff. The bulk sensitivity of the sensing structure used here was measured to be ⬃10 nm per refractive index unit by submerging the sensor in solutions of ethylene glycol and water. By varying the relative concentrations of ethylene glycol and water, refractive indexes between 1.33 共pure water兲 and 1.42 共90% ethylene glycol兲 were obtained.43 The refractive indexes of the SLB and buffer have previously been found to be ⬃1.48 and 1.335,22 respectively. Inserting these values in Eq. 共4兲 gives a decay length of ⬃5 nm. This corresponds to the decay length of the field extending from the silica surface. Note that the plasmonic field is strongest at the Au nanoparticle surface and decays exponentially from the latter. Thus, due to the relatively thick silica layer 共20 nm兲, only a minor part of the enhanced electromagnetic field is expected to extend from the silica surface and the relatively short decay length calculated here seems reasonable.44 C. Streptavidin binding „step II…

Exposure of the biotinylated bilayer to streptavidin 共step II兲 caused a decrease in ⌬f and a redshift of the plasmon resonance indicating that proteins bind to the bilayer. The dissipation increases during the protein binding, signaling the formation of a more viscoelastic layer. The equilibrium QCM-D responses for the streptavidin binding, ⌬f = ⬃ 39 Hz 共i.e., 690 ng/ cm2兲 and ⌬D = 1.3⫻ 10−6 are significantly larger than previously reported.9,45 The latter can most likely be attributed to a nonperfect bilayer on the nanostructured surface. POPC bilayers are in general protein resistant.46 However, if the bilayer is not perfect proteins may adsorb on the silica surface. The streptavidin binding serves as a good example of how the combined LSPR/QCM-D setup can be used to quantify the molecular mass of bound proteins from the LSPR peak shift and the wet mass measured with QCM.22 The only information required is the density 共or specific volume兲 of the proteins, which in most cases is well known. Assuming that all solvent in between the adsorbed molecules is coupled to the shear mode of the QCM oscillation 共which is a good approximation at high surface coverage and if the adsorbed entities form a rigid film兲 the thickness of the adsorbed layer

can be calculated from the Sauerbrey equation 关Eq. 共1兲兴 by using an effective film density described by38

␳film =

⌫QCM , 共5兲 兵共⌫LSPR/␳molecule兲 + 关共⌫QCM − ⌫LSPR兲/␳buffer兴其

where ⌫LSPR and ⌫QCM are the masses measured with LSPR and QCM respectively. ␳molecule and ␳buffer are the densities of the adsorbed molecules and the buffer solution. The densities of streptavidin and buffer were taken to be 1.3545 and 0.998,22 respectively. Inserting ␳film from Eq. 共5兲 into Eq. 共1兲 共the Sauerbrey equation兲 and solving for the film thickness t gives t = ⌫LSPR

冉

冊

1 1 ⌫QCM + − . ␳molecule ␳buffer ␳buffer

共6兲

If the film thickness t and the average refractive index of the film nfilm are known the adsorbed dry mass 共i.e., the mass measured with LSPR, ⌫LSPR兲 can be obtained from22,47 ⌫LSPR =

3t共nfilm − nbuffer兲共nfilm + nbuffer兲 2 共nfilm +

2 2 2兲关r共nbuffer + 2兲 − ␯共nbuffer − 1兲兴

,

共7兲

where r is the specific refractivity and ␯ is the specific volume 共i.e., the inverse of the density兲 of the adsorbed entities. The specific refractivity and the specific volume for streptavidin have previously been determined to be 0.249 and 0.719 cm3 / g, respectively.37 The average refractive index of the film can be determined from Eq. 共3兲 by replacing nSLB with nfilm, which gives nfilm =

共neff − nbuffer兲 + nbuffer , 1 − e共−t/L兲

共8兲

where neff is the effective refractive index within the sensing volume. neff can be calculated from Eq. 共4兲 knowing that the plasmon peak shift upon binding of streptavidin was ⌬␭ = 0.7 nm and estimating the bulk sensitivity to 8 nm per refractive index unit 共a 20% decrease in the bulk refractive index sensitivity has previously been observed after the formation of a SLB兲.22 Inserting ⌫LSPR from Eq. 共7兲 into Eq. 共6兲 and replacing nfilm with the expression in Eq. 共8兲, the following equation is obtained t

3共␣2 − nbuffer兲 2 2 − 1兲兴 共␣ + 2兲关r共nbuffer + 2兲 − ␯共nbuffer 2

⫻

冉

冊

1 1 ⌫QCM − −t= , ␳molecule ␳buffer ␳buffer

共9兲

where

␣=

neff − nbuffere共−t/L兲 . 1 − e共−t/L兲

By solving Eq. 共9兲 the thickness of the streptavidin layer is found to be 5.6 nm, which is a slightly larger thickness than has been observed previously.9,48–50 Using Eq. 共7兲 the adsorbed mass of streptavidin can finally be calculated to be ⬃370 ng/ cm2. As expected from the QCM-D measurement, this is a larger mass than what is usually observed for a layer of streptavidin, possibly due to the nonperfect SLB. However, it is interesting to compare the masses obtained with the

125105-7

Larsson et al.

two complementary techniques, 690 ng/ cm2 with QCM and 370 ng/ cm2 with LSPR. This indicates that the mass of bound solvent is 320 ng/ cm2 or 86% of the streptavidin mass, which is slightly lower than previously reported.9 D. Binding of biotinylated vesicles „step III…

Upon subsequent exposure to biotinylated vesicles 共step III兲 of the biotinylated bilayer with bound streptavidin, the frequency is observed to increase and the plasmon resonance redshifts monotonously until saturation. This, in combination with the high dissipation value 共⬎10⫻ 10−6兲, indicates the binding of intact vesicles. The frequency and dissipation shifts 共⌬f ⬇ 140 Hz and ⌬D ⬇ 12⫻ 10−6兲 are of the same magnitude as those reported previously for a similar vesicle layer.9 The magnitude of the plasmon peak shifts measured for the last two steps are approximately equal 共0.7 nm for streptavidin and 0.62 nm for vesicle binding兲 while the frequency shift for the vesicle binding is almost a factor four larger than that measured for streptavidin binding. The difference between the optical and QCM measurements clearly illustrates the difference between the two techniques. Vesicles contain a lot of trapped water, which is sensed by the QCM, while the optical technique measures the dry mass, which is much smaller. Another contributing factor to the relatively small plasmon shift observed for vesicle binding is that the gold nanoparticle sensitivity to changes in the surrounding dielectric environment decays exponentially with distance from the particle with a decay length of ⬃5 nm. The biotinylated vesicles bind to the streptavidin molecules which brings them ⬎10 nm away from the silica surface where the enhanced electromagnetic field and thus the sensitivity of the plasmonic structures is low. E. Gas phase measurements

To demonstrate the performance of the combined QCM/ LSPR equipment in a gas phase application, hydrogen absorption and release kinetics in palladium nanostructures were investigated. The latter is an important example of a gas phase reaction that benefits from simultaneous measurements with QCM and LSPR. Complete information about the storage kinetics and stored quantities cannot be obtained by QCM or LSPR alone but requires the combination of these two techniques.28 However, hydrogen absorption/desorption in palladium is only one example of a gas phase application that can benefit from the simultaneous measurements using QCM and LSPR. Other examples include metal oxidation/ corrosion and hydration/dehydration of e.g., hydrogels. Hydrogen potentially constitutes a very clean energy system and is an attractive energy carrier. It is nontoxic, carbon-free, abundant, and can generate energy either by combustion or 共with higher efficiency兲 in a fuel cell, producing electricity with only water as by-product. Two of the remaining technological obstacles, on the way to large-scale usage of hydrogen as a fuel, are the problems of clean and cost-efficient hydrogen production and safe, compact, and lightweight hydrogen storage. For the latter purpose, metal hydrides 共and also other hydrides like complex hydrides51兲 are interesting since they can reversibly store large amounts

Rev. Sci. Instrum. 80, 125105 共2009兲

of hydrogen gas. Solid state hydrogen storage is, therefore, proposed as a solution to circumvent the cost and safety concerns related to storage of gas phase hydrogen in, e.g., high pressure tanks.52 To remedy some of the major drawbacks of bulk hydride systems, i.e., slow absorption kinetics at temperatures of practical interest and the often high temperatures needed for sufficiently fast hydride decomposition, i.e., hydrogen release, nanostructured materials and nanoparticles have been proposed as a solution. Due to the altered thermodynamic and kinetic properties of nanosystems the latter are expected to ultimately lead to hydrogen storage systems with faster response and modified thermodynamics.53 Typically, when storing hydrogen in a metal, at low hydrogen concentration, hydrogen atoms form an interstitial dilute solid solution, the so-called ␣ phase. At higher hydrogen partial pressure 共i.e., when the hydrogen concentration increases at constant pressure兲, the metal hydride 共␤ phase兲 is formed, containing hydrogen atoms sitting in an ordered lattice in significant amounts.32 The phase transition from the ␣ phase to the ␤ phase is of first order, with a miscibility region where ␣ and ␤ phases coexist over a wide hydrogen concentration range 共referred to as “plateau” in a pressureconcentration 共p-C兲 isotherm兲. We have previously found that optical p-⌬␭max 共i.e., pressure-LSPR peak position兲 isotherms, for palladium disks, agree very well with p-C isotherms for the 共nanocrystalline兲 bulk palladium-hydrogen system, measured with other techniques.28 The latter implies a linear scaling of ⌬␭max with hydrogen concentration. To test that hypothesis, we performed separate QCM experiments to gravimetrically quantify the hydrogen concentration in the nanodisks. For that purpose, identical palladium nanodisk structures were prepared on one electrode of a stress compensated 共SC兲 cut quartz sensor crystal 共for details on these experiments the reader is referred to Ref. 54兲. The obtained QCM frequency shift during hydrogen exposure was related to the hydrogen concentration using the Sauerbrey equation 共Eq. 共1兲兲. The obtained QCM p-C isotherm, agreed very well with the corresponding LSPR-based one. Furthermore, the measured hydrogen/palladium ratios at different hydrogen pressures agreed very well with literature data for the nanocrystalline bulk palladium-hydrogen system. A linear scaling of the plasmonic ⌬␭max signal with hydrogen concentration was found. This demonstrates the application of QCM for calibration of LSPR-based optical hydrogen sensors. For such calibration purposes, it would be valuable to be able to carry out QCM and LSPR measurements simultaneously, on the same sample. Figure 4 shows hydrogen absorption and desorption kinetics in truncated palladium nanocones measured simultaneously by monitoring the frequency shift in the QCM and the optical extinction shift of the LSPR signal, using the combined LSPR/electrodeless QCM setup. The optical extinction was measured at a wavelength of 870 nm, which was on the short wavelength side of the peak maximum 共peak position ⬃1000 nm兲. The measured optical signal could, however, equally well have been that of the plasmon peak position. The extinction at 870 nm was chosen because in this particular measurement the plasmon peak maximum was

125105-8

Larsson et al.

FIG. 4. 共Color online兲 Optical extinction and QCM frequency shifts measured during hydrogen absorption in and desorption from truncated palladium cones. The hydrogen/palladium ratio calculated from the frequency shift, using the Sauerbrey equation, is also indicated in the figure.

outside the wavelength range reachable with the spectrometer. In this experiment, the hydride formation and decomposition kinetics at 22 ° C were monitored. Hydrogen was absorbed in palladium when flowing 4% H2 gas 共in Ar兲 over the crystal surface at atmospheric pressure. Changing the gas to 100% Ar caused hydrogen desorption. The sample was exposed to 4% hydrogen gas at t = 7 → 16 min, t = 28 → 39 min 10 s, and t = 53 min 30 s → t ⬎ 60 min. Absorption of hydrogen into the hydride phase increased the mass of the truncated cone structures, and thus caused the QCM frequency to decrease. Concurrently a decrease in the optical extinction was observed, corresponding to a spectral redshift of the plasmon resonance. The spectral redshift is a consequence of the changing dielectric properties of the palladium nanostructures upon hydrogen absorption.28,55,56 Changing to 100% Ar gas at t = 16 min and t = 39 min 10 s caused hydride decomposition and hydrogen desorption and thus the frequency and extinction increased to their original values. A hydrogen concentration of 4%, at atmospheric pressure, corresponds to a partial hydrogen pressure of 40.5 mbar, which is high enough to ensure total conversion of palladium to its hydride, as the corresponding hydrogen pressure is above the equilibrium plateau at 22 ° C.28 The frequency signal from the QCM and the optical LSPR response follow each other very closely, confirming again the linear dependence of the LSPR signal on the hydrogen concentration 共obtained by QCM兲 in the nanoparticle for the palladium-hydrogen system. The hydrogen concentration in the palladium nanostructures at saturation 共⌬f ⬇ 2.8 Hz兲 is 0.54 hydrogen atoms per palladium atom. The latter is in very good agreement with what has previously been observed both for nanocrystalline bulk and nanodisk palladium.54,57,58 The kinetics for hydrogen absorption were found to be faster than those for desorption 共Fig. 4兲, which is to be expected at 22 ° C and is related to different rate-limiting steps for the respective processes.59,60 In the absorption process the rate limiting step is connected with hydrogen diffusion, while it is governed by the final desorption step during hy-

Rev. Sci. Instrum. 80, 125105 共2009兲

drogen release. Using a blank quartz crystal, frequency shifts originating from differences in the temperature and viscosity of the two gases were found to be negligible at this relatively low hydrogen concentration. However, we would expect such effects to be more significant at higher hydrogen concentrations,61,62 which might make measurements of hydrogen storage kinetics, with QCM, difficult or even impossible. Such effects are also significant in batch type experiments.54 For storage kinetics experiments on more realistic hydrogen storage materials and nanoparticle systems, experiments of the type described above need to be carried out at higher hydrogen concentrations and temperatures, where QCM frequency perturbations caused by fluctuations in the viscosity or temperature can no longer be neglected. These unwanted contributions to the QCM signal significantly alter the appearance of the QCM kinetic curves obtained, rendering them very hard to interpret. However, if the magnitude of these effects is measured in a reference experiment on a blank crystal, it will still be possible to at least determine the total hydrogen concentration in the nanoparticles after the terminated hydriding reaction. On the contrary, the optically 共LSPR兲 measured kinetic curve is not affected by the viscosity or temperature fluctuations and, therefore, provides the complete and “real” kinetics information of interest. However, it does not contain quantitative information about the hydrogen concentration after the terminated reaction. Thus, by combining LSPR and QCM, the complete information of accurate kinetics and stored amount of hydrogen after terminated reaction can be obtained from the same sample simultaneously. F. The performance of the combined prototype system

For the liquid phase measurement, the short-term peakto-peak noise was found to be 0.15 Hz for QCM-D and 0.08 nm for the optical measurement, respectively. This corresponds to a mass resolution of 2.7 ng/ cm2 for QCM-D, if the Sauerbrey equation is used as reference, which is slightly larger than what is commonly obtained for individual QCM-D setups.20,63 Taking the formation of the lipid bilayer as an example, the optical resolution of 0.08 nm corresponds to a mass resolution of 42 ng/ cm2, which is low compared to what has previously been observed for other optical techniques.20,22 The optical resolution can most likely be improved by replacing the optically transparent PMMA, covering the hole in the top electrode, with a glass window, optimizing the light path through the cell, and by modifying the sensing structure to obtain a higher sensitivity to refractive index changes.44 Turning to the gas phase measurement of hydrogen absorption and desorption in palladium nanostructures, we note that the peak-to-peak noise is 0.10 Hz for QCM and 0.2% for the optical measurement, respectively. The corresponding mass resolution for QCM is 1.7 ng/ cm2. For the sensing structure used here, the QCM resolution corresponds to 0.02 hydrogen atoms per palladium atom, which can be compared with the optical resolution of 0.06 hydrogen atoms per palladium atom or 5.8 ng/ cm2. However, caution is required

125105-9

when comparing the sensitivities obtained with QCM and optical LSPR. Although the QCM sensitivity would increase drastically if the density of nanoparticles on the surface increased this is not the case for the resolution of the LSPR measurement. The optical measurement could in principle equally well be made on a single palladium nanoparticle. Although the signal-to-noise ratio in the optical measurement would decrease as the number of particles decreases, the effect is expected to be much smaller than for the QCM. The resolutions of the LSPR and QCM signals obtained here compare very well with similar measurements carried out using the individual respective techniques.28,54 VI. CONCLUDING REMARKS

An equipment for simultaneous measurement, on the same surface, using QCM共-D兲 and LSPR has been developed and its performance has been demonstrated both in gas and liquid phase. QCM共-D兲 was operated in an electrodeless configuration and optical access to the QCM crystal was obtained by making a hole in the middle of each of the two electrodes. Electrodeless QCM offers advantages in that it eliminates the need for fabrication of electrodes on the crystal surface, which is a consumable, and gives a larger freedom in choosing the surface functionalization. It also provides maximum freedom for preparation of the LSPR sensing structure. Although the information obtained from LSPR measurements is similar to that obtained using other optical techniques, e.g., reflectometry and SPR, major advantages of the present setup are the much simpler optical equipment needed, much less critical beam alignment, and a larger freedom in surface structuring/functionalization. We have shown that the results obtained using the electrodeless QCM共-D兲 are closely similar to those obtained using regular QCM共-D兲 measurements, and that valuable, complementary information can be obtained by combining QCM共-D兲 with LSPR-based sensing in gas and liquid environments. ACKNOWLEDGMENTS

The PhotoNano and NanoX programs, funded by the Swedish Foundation for Strategic Research, and the Swedish Research Council 共Grant No. 16254111兲, are gratefully acknowledged for financial support. J. F. Alder and J. J. MacCallum, Analyst 共Cambridge, U.K.兲 108, 1169 共1983兲. 2 G. Sauerbrey, Z. Phys. 155, 206 共1959兲. 3 M. Rodahl, F. Höök, A. Krozer, P. Brzezinski, and B. Kasemo, Rev. Sci. Instrum. 66, 3924 共1995兲. 4 M. Rodahl and B. Kasemo, Rev. Sci. Instrum. 67, 3238 共1996兲. 5 C. Kepplinger, N. Ilk, U. B. Sleytr, and B. Schuster, Soft Matter 5, 325 共2009兲. 6 D. Thid, K. Holm, P. S. Eriksson, J. Ekeroth, B. Kasemo, and J. Gold, J. Biomed. Mater. Res. Part A 84A, 940 共2008兲. 7 M. A. Plunkett, Z. Wang, M. W. Rutland, and D. Johannsmann, Langmuir 19, 6837 共2003兲. 8 J. Zhou, S. Moya, L. Ma, C. Gao, and J. Shen, Macromol. Biosci. 9, 326 共2009兲. 9 M. Edvardsson, S. Svedhem, G. Wang, R. Richter, M. Rodahl, and B. Kasemo, Anal. Chem. 81, 349 共2009兲. 10 S. Löfås, M. Malmqvist, I. Rónnberg, E. Stenberg, B. Liedberg, and I. Lundström, Sens. Actuators B 5, 79 共2009兲. 11 H. Arwin, Thin Solid Films 377, 48 共2000兲. 1

Rev. Sci. Instrum. 80, 125105 共2009兲

Larsson et al. 12

J. C. Dijt, M. A. Cohen Stuart, and G. J. Fleer, Adv. Colloid Interface Sci. 50, 79 共1994兲. 13 C. Hänel and C. Gauglitz, Anal. Bioanal. Chem. 372, 91 共2002兲. 14 J. Vörös, J. J. Ramsden, G. Csucs, I. Szendro, S. M. De Paul, M. Textor, and N. D. Spencer, Biomaterials 23, 3699 共2002兲. 15 M. Rodahl, F. Höök, C. Fredriksson, C. A. Keller, A. Krozer, P. Brzezinski, M. V. Voinova, and B. Kasemo, Faraday Discuss. 107, 229 共1997兲. 16 L. E. Bailey, D. Kambhampati, K. K. Kanazawa, W. Knoll, and C. W. Frank, Langmuir 18, 479 共2002兲. 17 A. Laschitsch, B. Menges, and D. Johannsmann, Appl. Phys. Lett. 77, 2252 共2000兲. 18 J. Rishpon, A. Redondo, C. Derouin, and S. Gotttesfeld, J. Electroanal. Chem. 294, 73 共1990兲. 19 T. Tjärnhage and G. Puu, Colloids Surf., B 8, 39 共1996兲. 20 G. Wang, M. Rodahl, M. Edvardsson, S. Svedhem, G. Ohlsson, F. Höök, and B. Kasemo, Rev. Sci. Instrum. 79, 075107 共2008兲. 21 A. B. Dahlin, P. Jönsson, M. P. Jonsson, E. Schmid, Y. Zhou, and F. Höök, ACS Nano 2, 2174 共2008兲. 22 M. P. Jonsson, P. Jönsson, and F. Höök, Anal. Chem. 80, 7988 共2008兲. 23 T. Nomura, F. Tanaka, T. Yamada, and H. Itoh, Anal. Chim. Acta 243, 273 共1991兲. 24 O. Wolff, E. Seydel, and D. Johannsmann, Faraday Discuss. 107, 91 共1997兲. 25 G. H. Chan, J. Zhao, E. M. Hicks, G. C. Schatz, and R. P. Van Duyne, Nano Lett. 7, 1947 共2007兲. 26 C. Langhammer, M. Schwind, B. Kasemo, and I. Zoric, Nano Lett. 8, 1461 共2008兲. 27 E. M. Larsson, C. Langhammer, I. Zoric, and B. Kasemo, Science 326, 1091 共2009兲. 28 C. Langhammer, I. Zoric, B. Kasemo, and B. M. Clemens, Nano Lett. 7, 3122 共2007兲. 29 J. N. Anker, W. P. Hall, O. Lyandres, N. C. Shah, J. Zhao, and R. P. Van Duyne, Nature Mater. 7, 442 共2008兲. 30 A. J. Haes, L. Chang, W. L. Klein, and R. P. Van Duyne, J. Am. Chem. Soc. 127, 2264 共2005兲. 31 I. H. El-Sayed, X. H. Huang, and M. A. El-Sayed, Nano Lett. 5, 829 共2005兲. 32 Y. Fukai, The Metal-Hydrogen System 共Springer, Berlin, 1993兲. 33 H. Fredriksson, Y. Alaverdyan, A. Dmitriev, C. Langhammer, D. S. Sutherland, M. Zäch, and B. Kasemo, Adv. Mater. 19, 4297 共2007兲. 34 P. Hanarp, M. Kall, and D. S. Sutherland, J. Phys. Chem. B 107, 5768 共2003兲. 35 R. P. Richter, R. Bérat, and A. Brisson, Langmuir 22, 3497 共2006兲. 36 R. Richter, A. Mukhopadhyay, and A. Brisson, Biophys. J. 85, 3035 共2003兲. 37 E. Reimhult, C. Larsson, B. Kasemo, and F. Höök, Anal. Chem. 76, 7211 共2004兲. 38 C. A. Keller and B. Kasemo, Biophys. J. 75, 1397 共1998兲. 39 A. I. Greenwood, S. Tristam-Nagle, and J. F. Nagle, Chem. Phys. Lipids 143, 1 共2006兲. 40 A. Mashaghi, M. Swann, J. Popplewell, M. Textor, and E. Reimhult, Anal. Chem. 80, 3666 共2008兲. 41 Z. Salamon and G. Tollin, Biophys. J. 80, 1557 共2001兲. 42 C. A. Keller, K. Glasmastar, V. P. Zhdanov, and B. Kasemo, Phys. Rev. Lett. 84, 5443 共2000兲. 43 E. M. Larsson, J. Alegret, M. Käll, and D. S. Sutherland, Nano Lett. 7, 1256 共2007兲. 44 T. Rindzevicius, Y. Alaverdyan, M. Käll, W. A. Murray, and W. L. Barnes, J. Phys. Chem. C 111, 11806 共2007兲. 45 C. Larsson, M. Rodahl, and F. Höök, Anal. Chem. 75, 5080 共2003兲. 46 K. Glasmästar, C. Larsson, F. Höök, and B. Kasemo, J. Colloid Interface Sci. 246, 40 共2002兲. 47 J. A. De Feijter, J. Benjamins, and F. A. Veer, Biopolymers 17, 1759 共1978兲. 48 T. C. Edwards, S. Koppenol, W. Frey, W. R. Schief, Jr., V. Vogel, R. E. Stenkamp, and P. S. Stayton, Langmuir 14, 4683 共1998兲. 49 W. A. Hendrickson, A. Pahler, J. L. Smith, Y. Satow, E. A. Merritt, and R. Paul Phizackerley, Proc. Natl. Acad. Sci. U.S.A. 86, 2190 共1989兲. 50 I. Reviakine and A. Brisson, Langmuir 17, 8293 共2001兲. 51 S. I. Orimo, Y. Nakamori, J. R. Eliseo, A. Züttel, and C. M. Jensen, Chem. Rev. 共Washington, D.C.兲 107, 4111 共2007兲. 52 L. Schlapbach and A. Zuttel, Nature 共London兲 414, 353 共2001兲. 53 V. Bèrubè, G. Radtke, M. Dresselhaus, and G. Chen, Int. J. Energy Res.

125105-10

31, 637 共2007兲. C. Langhammer, E. M. Larsson, I. Zorić, and B. Kasemo, 共unpublished兲. 55 J. Isidorsson, I. A. M. E. Giebels, H. Arwin, and R. Griessen, Phys. Rev. B 68, 115112 共2003兲. 56 A. C. Switendick, J. Less-Common Met. 130, 249 共1987兲. 57 T. Mutschele and R. Kirchheim, Scr. Metall. 21, 1101 共1987兲. 58 T. Mutschele and R. Kirchheim, Scr. Metall. 21, 135 共1987兲. 54

Rev. Sci. Instrum. 80, 125105 共2009兲

Larsson et al.

H. Conrad, G. Ertl, and E. E. Latta, Surf. Sci. 41, 435 共1974兲. V. P. Zhdanov and B. Kasemo, Chem. Phys. Lett. 460, 158 共2008兲. G. A. Frazier and R. Glosser, J. Less Common Met. 74, 89 共1980兲. 62 I. Kulchytskyy, M. G. Kocanda, and T. Xu, Appl. Phys. Lett. 91, 113507 共2007兲. 63 G. M. L. Messina, C. Satriano, and G. Marletta, Colloids Surf., B 70, 76 共2009兲. 59 60 61