Human Serum Albumin Adsorption Study on62-MHz Miniaturized Quartz Gravimetric Sensors

Ping Kao,† Ashish Patwardhan,† David Allara,*,‡,§,| and Srinivas Tadigadapa*,†,|

Department of Electrical Engineering, Department of Chemistry, Materials Science & Engineering, and the MaterialsResearch Institute, Pennsylvania State University, University Park, Pennsylvania 16802

We have designed and fabricated 25-µm-thick quartzresonators operating at a fundamental resonance fre-quency of ∼62 MHz. The results show a substantialincrease in the mass sensitivity compared to singlemonolithic commercial resonators operating at lowerfrequencies in the ∼5-10-MHz range. The overall per-formance of the micromachined resonators is demon-strated for the example of human serum albumin proteinadsorption from aqueous buffer solutions onto goldelectrodes functionalized with self-assembled monolayers.The results show a saturation adsorption frequency changeof 6.8 kHz as opposed to 40 Hz for a commercial ∼5-MHz sensor under identical loading conditions. From theanalysis of the adsorption isotherm, the equilibriumadsorption constant of the adsorption of the protein layerwas found to be K ) 8.03 × 10^6 M^-1, which is inagreement with the values reported in the literature. Thehigh sensitivity of the miniaturized QCM devices can bea significant advantage in both vapor and solution adsorp-tion analyses.

Commercial quartz crystal microbalances (QCM) have beenextensively used for the characterization and measurement ofdeposited and adsorbed thin films in both vacuum and liquidenvironments.1–4 The most commonly available commercial QCMconsists of a ∼300-µm-thick and ∼25-mm-diameter disk made froman AT-cut quartz crystal with gold electrodes on each face, has ashear resonance frequency in the 5-MHz range, and exhibits asensitivity of ∼17 ng cm-2Hz-1.5 Although quartz crystals withfundamental resonances in the 10-MHz range and a few with evenhigher frequencies are commercially available, the inherentfragility of large-area thinner resonators has precluded theirextensive use in comparison to the 5-MHz resonators.6 Under theconditions that the adsorbed film material is rigid, sufficiently thincompared to the quartz crystal, and attached to the sensor surface

under a no-slip condition, the dependence of the frequency change(∆f) of a resonating quartz crystal to the mass loading (∆m) isgiven quite accurately by the Sauerbrey equation:

∆f)-(f0 ⁄AxqFq)∆m)-(2f02 ⁄ √µqFqA)∆m)-Sf

∆mA

(1)

where f0 is the fundamental resonant frequency with no attachedmass, µq is the shear modulus of the quartz (2.947 × 1010 N m-2

for AT-cut quartz), Fq is the density of quartz (2.648 × 103 kg m-3),xq is the thickness of the crystal, and A is the area of the electrodeon the quartz crystal. The negative sign indicates a reduction inthe resonance frequency upon mass loading. A very importantaspect of eq 1) is that the change in frequency (∆f) for the samemass loading (∆m) increases as the thickness xq and the area Aof the crystal are made smaller. This provides a possible way ofmaking large enhancements in the performance of quartz crystalresonators if methods of fabrication can be developed to miniatur-ize them. The purpose of the work reported in this paper is todemonstrate that such improvements are possible.

One of the important applications of QCM devices is to assayadsorption in liquid environments, often with significant viscosity,and these effects were of considerable interest in our developmentof miniature resonators. These conditions can both compromisethe ultimate performance of the devices by damping of theresonance and change the form of the frequency-mass loadingresponse. Examples include polymer adsorption, electrochemicaldeposition/stripping, protein adsorption, and cell attachment/-detachment.7,8 It is important to assess these effects in thedevelopment of new types of QCMs. In such applications, thefundamental mode QCM frequency shift in eq (1) is modified toone that depends upon the square root of the viscosity-densityproduct of the liquid, as given by

∆f)-f0

3⁄2

√πFqµq√FLηL (2)

where FL and ηL are the density and viscosity of the liquid,respectively.1 In liquid medium applications, the shear waverapidly damps out as it travels through the thickness of the liquid,and consequently, the QCM typically samples a layer of thickness

* To whom correspondence should be addressed. E-mail: sat10@psu.edu; dla3@psu.edu.

† Department of Electrical Engineering.‡ Department of Chemistry.§ Materials Science & Engineering.| Materials Research Institute.

(1) Kanazawa, K. K.; Gordon, J. G. Anal. Chem. 1985, 57 (8), 1770–1771.(2) Janshoff, A.; Galla, H.-J.; Steinem, C. Angew. Chem., Int. Ed. 2000, 39,

4004–4032.(3) Granstaff, V. E.; Martin, S. J. J. Appl. Phys. 1994, 75 (3), 1319.(4) Schumacher, R. Angew. Chem., Int. Ed. Engl. 1990, 29 (4), 329–343.(5) O’Sullivan, C. K.; Guilbault, G. G. Biosens. Bioelectron. 1999, 14 (8-9),

663–670.

(6) The 9-10-MHz quartz resonators are commercially available from MaxtekInc. (a division of Inficon), and 27-MHz resonators are commerciallyavailable from Initium Inc. (a division of Ulvac).

(7) Kanazawa, K. K.; Melroy, O. R. IBM J. Res. Dev. 1993, 37 (2), 157.(8) Bruckenstein, S.; Shay, M. J. Electroanal. Chem. 1985, 188 (1-2), 131.

Anal. Chem. 2008, 80, 5930–5936

10.1021/ac8005395 CCC: $40.75 2008 American Chemical Society5930 Analytical Chemistry, Vol. 80, No. 15, August 1, 2008Published on Web 06/21/2008

equivalent to the decay length δ = (ηL/πFLf0)0.5. The change inthe dissipation factor can be written as9

∆D)2f0

1⁄2

√πFqµq√FLηL (3)

For a commercial 5-MHz QCM, the typical decay length inwater is ∼250 nm, which is large in comparison to the thicknessof molecular or even polymer and biomolecular (e.g., proteins)films, which often can reach thickness of 10-50 nm. As a result,a commercial QCM not only samples the adsorbed film but isalso significantly affected by the viscous fluid layer above it.However, if the QCM thickness is decreased, its fundamentalresonance frequency increases inversely as a function of thethickness of the quartz and thus the decay length in liquid alsodecreases consequently. For example, if the thickness of a quartzresonator is decreased from 330 (commercial 5-MHz resonator)to 25 µm (miniaturized 62-MHz resonator), the decay length inwater is reduced to ∼68 nm.

Motivated by the fact that the sensitivity to mass loading andthe interfacial depth response can be significantly improved viaminiaturization of quartz resonators, we have designed andfabricated 25-µm-thick devices operating at a fundamental reso-nance frequency of ∼62 MHz. In this paper, we present theperformance of these resonators using the example of adsorptionof human serum albumin (HSA) protein from buffered solutiononto gold electrodes functionalized with self-assembled monolay-ers. While single resonator devices with fundamental resonancefrequencies up to 94 MHz have been reported, the behavior ofresonators with fundamental frequencies higher than 30 MHz,under liquid and viscoelastic loading conditions, has not beenstudied or reported thus far.10–13 Such a study is imperative forthe development of miniaturized QCM-based sensor arrays forapplications such as biochemical sensing. This study helps to fillthis gap and demonstrates that much higher performance QCMdevices than currently available are possible through appropriatefabrication and engineering.

EXPERIMENTAL SECTIONMiniaturized Resonator Design Considerations. The pri-

mary challenge is to be able to maintain a high Q-factor whilesimultaneously suppressing spurious resonance modes. A resona-tor with an electroded area of limited extent (radius rq) on a finiteradius wafer that in turn is mounted on low Q supports can havea high Q-factor only if a very large fraction of its vibration energyis restricted to the electrode regions. Energy trapping in the caseof such finite-sized electrode resonators is dominated by totalinternal reflection of vibrations at the edge of the electrodes. Incases for which the width of the electrodes is small compared to

the wafer thickness, leakage of acoustic energy through thevolume of the wafer increases considerably and vibrational energyreflection at the electrode edges becomes increasingly small. Asa consequence, poor Q-factors are observed.

On the other hand, for the effective suppression of the spuriousmodes, the rule of thumb thickness of quartz to electrode radiusratio is given by14

∆fpb < f02.337

N2 (xq

rq) (4)

where f0 is the resonance frequency of the quartz crystal ofthickness xq after the mass loading due to the electrodes, and∆fpb is the plate back frequency i.e., the frequency differencebetween the resonance frequency of the bare crystal and afterdeposition of top and bottom electrodes of radius rq. This impliesthat as the quartz resonator thickness is reduced the electrodediameter also must be reduced in order to effectively suppressspurious modes. Excessive decrease in the electrode diameter,however, will result in acoustic energy leakage and poor confine-ment of energy resulting in small Q-factors.

These opposing requirements entail a design compromise. Ourresults show that, for 50-70-MHz resonators with 100 nm of goldelectrodes on both faces, the optimal electrode diameter (2rq) toresonator thickness (xq) ratio is ∼10-20. Large deviations fromthis value either way result in resonators with poor characteristics.One important consideration to note is that, once etched, a 25-µm-thick quartz crystal is extremely fragile to handle and readilyshatters. Thus, the thinned resonator regions need to be es-sentially carved into a thicker quartz plate. This coupled with therequirement to allow for biochemical (liquid) experiments to beperformed on the resonator surface results in an inverted mesadesign, shown in Figure 1, in which the thin resonator regionsare etched out of a thicker quartz plate. The thicker regions allowfor easy handling of the device, and these regions are essentiallyglued onto a package to allow for the liquid testing. To allow foreasy contact to the electrode on the etched side, the etchedregions of the pixels were extended to the edge of the chip asshown in Figure 1. The unetched surface was designed to be thereaction surface and had the “front-side” electrode. This config-

(9) Rodahl, M.; Kasemo, B. Sens. Actuators, B 1996, B37 (1-2), 111.(10) Abe, T.; Esashi, M. Sens. Actuators, A 2000, A82 (1-3), 139–143.(11) Abe, T.; et al. Energy dissipation in small-diameter quartz crystal microbal-

ance experimentally studied for ultra-high sensitive gravimetry. In MicroElectro Mechanical Systems; MEMS-03 Kyoto, IEEE The Sixteenth AnnualInternational Conference, 2003.

(12) Buttgenbach, S. Proc. SPIE-Int. Soc. Opt. Eng. 2001, 4205, 207–217.(13) Rabe, J. B. S.; Schroder, J.; Hauptmann, P. Monolithic miniaturized quartz

microbalance array and its application to chemical sensor systems forliquids. In Sensors 2002; Proc. IEEE, 2002.

(14) Lucklum, R.; Eichelbaum, F. Interface circuits for QCM sensors. InPiezoelectric Sensors; Steinem, C., Janshoff, A., Eds.; Springer Verlag: NewYork, 2007; pp 3-47.

Figure 1. Schematic illustration of the micromachined quartz crystalmicrobalance.

5931Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

uration allowed the top (reaction)-side electrode to be at lowpotential while the resonator was excited via the “back-side”electrode.

Fabrication of the Quartz Crystal Microbalance. An induc-tively coupled plasma (ICP) etch process was used for quartzetching.15 The process uses a mixture of SF6 and Ar as the etchinggases. Mirror finish with average surface roughness less than 2nm was achieved even after an etch depth of 75 µm by etching ata very low base pressure of 1 mTorr. Electroplated nickel on aAu/Cr seed layer was used as the etching mask.

The fabrication process of the quartz crystal microbalance isillustrated schematically in Figure 2. A patterned nickel hard masklayer was deposited by electroplating (Figure 2b). The quartzcrystal was thereafter etched in a high-density ICP RIE etcherusing SF6 + Ar gases (Figure 2c). High selectivity (∼10) for theNi mask and high etch rates (∼0.5 µm/min) using an ICP etcherhave been already reported by our and other groups.15,16 The hardnickel mask was then stripped and lithographic patterning andetching performed on the bottom side Cr/Au electrodes (Figure2d). Finally the top-side electrodes were aligned and patternedusing liftoff to complete the QCM device (Figure 2e).

Figure 2f shows an optical picture of the fabricated QCM. A 5× 5 mm2 square hole was machined in a 24-pin, dual in-line ceramicpackage (DIP). The plane (unetched) side of the resonator havingthe front-side electrode where the mass sensing experiments areto be performed was placed facing the machined hole in thepackage and attached using room-temperature vulcanized siliconeelastomer adhesive. The front-side electrode was electricallyconnected to the gold layer on the DIP package using silver epoxyand wire bonded to one of the pads while the back-side electrodeof the resonator was directly wire bonded to any other availablepad in the package. An Agilent 4294A impedance analyzer wasused for all the reported characterization of the resonance

frequency and dissipation. All measurements reported in this workare based on single pixel measurements (details on parallelmeasurements will be reported elsewhere). Tested resonatorswere ∼25 µm thick, electrode diameter of 500 µm, with afundamental resonance frequency of ∼62 MHz.

Experimental Procedure and Setup. Materials. HSA (SigmaAldrich) and glycerol (JT Baker) were used directly. Thephosphate buffer solution (PBS) was made by dissolving phosphate-buffered saline powder (Sigma Aldrich) into 18 MΩ · cm DI water(Millipore Milli-Q system; Barnstead International). Different HSAconcentration solutions were made by diluting a 10 mg/mL stocksolution. Water-glycerol mixtures, ranging from pure water to80% glycerol, were made in increments of 10%.

Surface Functionalization. The micromachined QCM was cleanedby three cycles of exposure to UV ozone, each 30 min long, followedby thorough rinsing with ethanol and immersion in ethanol for 1 h.Finally, the electrodes were exposed to 1 mM hexadecanethiolsolution for 48 h. This procedure has been demonstrated in ourlaboratory to result in the formation of a hexadecanethiolate self-assembled monolayer (HD-SAM), which is well organized and highlyhydrophobic (as verified by advancing contact angles of ∼115 and46° for water and hexadecane, respectively).

Instrument Setup. An Agilent 4294A impedance analyzer was usedto characterize the micro-QCM device. Ceramic dual in-line packagessimilar to the ones used for the packaging of the QCM device weremodified to provide open and short circuit compensation fixturesfor accurate calibration of the QCM. A 100 Ω resistor was used tosimulate the load for compensation at 62-MHz frequency range. Theimpedance analyzer was set up to simultaneously measure theimpedance magnitude and phase angle as a function of frequency.The calibrated impedance and phase angle data were recorded usingLabview data acquisition setup. A nonlinear regression was used tofit the phase angle data to a Lorentzian function. All measurementswere carried out inside a 2 in. × 3 in. × 5 in. aluminum die-cast boxto prevent rf interference from the surroundings. The packaged QCMdevice was suspended from two BNC connectors fixed on the boxusing soldered wires. Finally, the custom-made die-cast box wasplaced inside a large, temperature-controlled chamber with thetemperature set at 23(±0.1) °C.

Protein Adsorption. Prior to performing protein adsorptionexperiments, the gold electrode surfaces were functionalized withthe HD-SAM. A ∼15-µL drop of PBS solution was delivered tothe electrode surface, and the QCM was allowed to stabilize for1 h to a constant frequency. This was followed by the sequentialinjection of increasing concentration protein solutions. Prior tochanging to each new concentration solution, the QCM surfacewas rinsed twice with PBS solution, rinsed thoroughly, and gentlydried under a filtered N2 stream.

RESULTSUnder the temperature-controlled conditions of 23 ± 0.1 °C,

the as-fabricated resonators exhibited a fundamental resonancefrequency of ∼62 MHz with frequency stability (drift) of around30 Hz over a period of 10 h and a frequency fluctuation noise of∼13 Hz in air. The Q-factor of the resonators was found to be∼6758 with a phase rotation of ∼10°. For the same resonator(bare, i.e., without SAM layer) in PBS solution, the Q-factordecreased to 3396 with a frequency noise of ∼25 Hz measuredover a period of 30 min. Coating the resonator with the SAM layer

(15) Goyal, A.; Hood, V.; Tadigadapa, S. J. Non-Cryst. Solids 2006, 352 (6-7),657.

(16) Abe, T.; Esashi, M. Sens. Actuators, A 2000, 82, 139–143.

Figure 2. (a-e) Schematic illustration of the fabrication processused. See text for detailed description of the individual steps. (f)Optical photo of a fabricated QCM (g) Packaged sensor in a modifieddual-in line ceramic package with a plastic reactor attached usingsilicone adhesive.

5932 Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

had little effect on the frequency noise or drift, which increasedby ∼1% of the values for the bare resonator.

Calibration Experiments. Two separate experiments werecarried out to measure the mass loading-frequency sensitivityof the devices in order to compare with the predictions from eqs(1) and (2). In the first measurement, a QCM device with afundamental frequency of 62.319 367 MHz, was placed inside avacuum deposition chamber with the sample connected to thefrequency measurement via an electrical feedthrough. Depositionof a 3.0-nm-thick, highly uniform gold film resulted in ∆f )-10.524 ± 0.025 kHz, or –3.51 kHz ·nm-1. In contrast, eq (1)predicts ∆f ) -51.684 kHz, or –17.2 kHz ·nm-1. Thus, thesensitivity of the fabricated micromachined QCM was found tobe 0.203 (∼1/5) of the theoretical value as computed using eq 1.

The second calibration was done to check the dependence of∆f on medium viscosity as determined from water-glycerolmixtures. As seen in Figure 3, except at very high concentrationsof 50% glycerol and above, a plot ofFLηL versus ∆f and dissipation(∆D) change shows a near-linear dependence, as expectedaccording to eqs 2 and 3, respectively. Once again, however, thepredicted ∆f by eq 2 is found to be ∼6 times larger than themeasured ∆f. Dividing eq 2 by eq 3 we obtain

-2 ∆f∆D

) f0 (5)

Note the experimental ratio of ∆f/∆D ∼ -f0/2, as predicted byeq 5.

These calibration experiments show that the frequency sen-sitivity of the fabricated QCM is ∼0.17-0.2 of the theoreticallyexpected value. It is important to note that the change indissipation factor also reduces by a proportional amount, establish-ing a self-consistency in the resonance frequency value. Althoughthe reduced, sensitivity behavior is not clearly understood at thistime, it is important to emphasize that the absolute magnitude ofthe frequency change, as well as the signal-to-noise ratio, for theseminiaturized resonators is superior to commercially available5-MHz QCMs. Table 1 summarizes the performance of theminiaturized QCM under various glycerol loading conditions.

Protein Adsorption Experiment. General Adsorption IsothermShape and Dissipation Characteristics of the HSA Layer. A plot ofthe resonance frequency shifts versus HSA solution concentration

for adsorption on HD-SAM-coated gold electrodes is shown inFigure 4. A similar shaped curve was obtained in an earlier studyusing a nominal 5-MHz QCM.17 Both sets of data show very closefits with high (>0.99) R2 values to a sigmoidal function. The curvesdiffer, however, in that the full span of the frequency shift oversimilar HSA concentration ranges was ∼40 Hz for the 5-MHzQCM in contrast to the ∼6.8-kHz span for our 62-MHz QCM(Figure 4). For saturation coverage, this implies a signal-to-noiseratio of 523 in air and 272 in PBS with an insignificantly smalldecrease for the SAM-coated device. These values are ∼20 timeshigher than the signal-to-noise ratio of a 5-MHz resonator.

In Figure 5, the observed change in the dissipation factor ∆Dis plotted against the change in frequency ∆f for the variousconcentrations of HSA. The slope reveals the rigidity of theadsorbed layer. A low value of ∆D/∆f indicates a highly rigid,low-dissipation layer, and conversely, a high ∆D/∆f value indicatesa soft, dissipative film. From Figure 5, the adsorption of HSA onthe HD-SAM surface shows ∆D/∆f of ∼7.28 × 10-9 s, a low valueconsistent with a rigid adsorbed film.18 Previous work17 has shownthat the HSA protein forms a monolayer near saturation coverages,and given the relatively small size of HSA (66.3 kDa) incomparison to other proteins, the observed high rigidity of theadsorbed HSA film is not surprising.

Langmuir Isotherm Model Behavior. Assuming a limitingcoverage of a single, uniform layer, the HSA adsorption behaviorcan be tested for a fit to a Langmuir isotherm. The basic Langmuirequilibrium is represented as

A+ * 98K

A* (6)

where A is the solution species, * represents an adsorbate-freeregion of the surface, A* is an adsorbate species, and K is theassociated equilibrium constant. The Langmuir isotherm modelassumes (i) a reversible adsorption process, (ii) no lateralinteractions between adsorbates, and (iii) a single uniformadsorbate layer at saturation coverage. At equilibrium, the cover-age is independent of time and can be given by

Γeq

Γmax) KC

1+KC(7)

where Γeq and Γmax are the number of adsorbates per unit area atthe equilibrium point for solution concentration C and at thesaturation coverage (asymptotically approached as Cf ∞). In theactual HSA adsorption experiment, Γeq and Γmax are representedby the masses per unit area (∆m and ∆mmax) of adsorbed HSA ateach particular HSA concentration (C) and at the maximum limit,respectively. In turn, ∆m and ∆mmax are determined directly fromthe observed frequency shifts (∆f and ∆fmax). Equation (7) cannow be rewritten as

C∆m

) C∆mmax

+ 1(K∆mmax)

(8)

From the plot of C/∆f versus C in

(17) Krishnan, A.; Liu, Y.-H.; Cha, P.; Allara, D. L.; Vogler, E. A. J. R. Soc. Interface2006, 3 (7), 283.

(18) Hook, F.; Rodahl, M.; Brzezinski, P.; Kasemo, B. Langmuir 1998, 14 (4),729–734.

Figure 3. Decrease in the resonance frequency and the increasein the dissipation factor as function of weight percent concentrationof glycerol in DI water.

5933Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

Calculation of the Langmuir equilibrium constant over the fullrange of values yields a value of K ) 8.03 × 106 M-1. As a furthertest of the validity of the Langmuir model, the data can be fit atthe low-concentration end where the adsorbate molecules arestatistically isolated from one another, minimizing adsorbate-adsorbate interaction effects and mass transport limited adsorp-tion-desorption kinetic deviations from true equilibrium. Fittinga straight line to the first five data points at low concentrations(Figure 6) yields K ∼3.5 × 108 M-1, a value ∼43 times largerthan the full concentration range value above. It must be notedthat in the low-concentration region there can be considerable

difference between the initial and equilibrium concentrations ofHSA since a sizable proportion of it can be adsorbed out of thesolution phase on to the QCM surface. Since we did notindependently confirm the equilibrium concentration of HSA inthe supernatant liquid, we use the initial concentration values inthe generation of this plot. This can explain some of the differenceobserved between the K-values obtained in the high- and low-concentration regimes. These values of K span the intermediatevalue of K ) 2.12 × 107 M-1 reported for HSA adsorption on silicaparticles.19 Furthermore, the differences at both ends can beattributed to differences in the experimental conditions and thechemical natures of the surfaces in the two studies. Whereas inour work the protein solution is essentially allowed to equilibrateunder “static” conditions, in the earlier report the interactionbetween the silica particles and protein was enhanced by activeflow conditions. With regard to chemical differences, our studiesinvolve a highly hydrophobic surface (HD-SAM) in contrast tothe hydrophilic SiO2 surface in the earlier report.

Viscoelastic Characteristics of the Adsorbed HSA Layer. Foranalysis of viscoelastic effects, the adsorbed protein layer adjacentto the solution medium can be considered schematically as shownin Figure 7. Given the thin (∼2 nm), chemically attached, densenature of the HD-SAM layer, it can be accurately considered as arigid extension of the electrode. The adsorbed HSA layer on theother hand is only attached to the HD-SAM surface via van derWaal’s forces with each molecule surrounded by water and in

(19) Docoslis, A.; Wu, W.; Giese, R. F.; van Oss, C. J. Colloids Surf., B:Biointerfaces 1999, 13 (2), 83.

Table 1. Glycerol Loading Performance of the Micro-QCM at the Fundamental Resonance Modea

wt % fmeas (MHz) ∆fcalc* (Hz) ∆fmeas (Hz) Dmeas (10-4) ∆Dmeas (10-4) f ) -2∆fmeas/∆Dmeas (MHz) f(calc)/f(meas)

10 62.672 279 36792.9 5381 3.6721 1.90 56.72 0.9120 62.365747 6 43254.6 6861 4.1581 2.38 57.58 0.9230 62.365 163 6 52060.9 7445 4.4673 2.69 55.30 0.8940 62.363 583 7 64534.1 9025 5.0177 3.24 55.66 0.8950 62.361 594 7 82969.8 11014 5.8126 4.04 54.55 0.8760 62.358 225 6 113028.0 14383 6.9679 5.19 55.39 0.8970 62.350 222 7 165459.7 22386 10.015 8.24 54.33 0.8780 62.343 263 2 275127.9 29345 12.548 10.8 54.48 0.87

a The change in frequency ∆(fcalc) has been calculated using eq (2). f0(air) ) 62.372 608 6 MHz and D0(air)) 1.7748 × 10-4 have been used tocalculate the ∆fmeas and ∆Dmeas.

Figure 4. Shift of the QCM resonance frequency as a function ofthe natural logarithm of the concentration of HSA in solution foradsorption on a HD-SAM gold electrode.

Figure 5. Change in the dissipation factor (∆D) plotted against thechange in the resonance frequency (∆f) for different concentrationsof HSA protein solutions. The slope of the linear fit yields a value of7.8 × 10-9 s.

Figure 6. Langmuir model plot of adsorbed HSA mass, representedby ∆f, versus HSA concentration in the high concentration range. Thesolid line is a linear fit to the data, consistent with a Langmuir behavior.The inset shows the slope.

5934 Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

contact with the PBS.17 In this situation, the HSA layer isconsidered to have some degree of viscoelastic character withthe surrounding PBS solution, containing only a very low volumefraction of protein molecules, represented as a Newtonian fluid.

According to Krishnan and co-workers, an adsorbed HSA layerat saturation coverage on a hydrophobic surface can be modeledas having a hexagonal type of close packing, a coverage of 750ng · cm-2 and an ∼7-10-nm thickness.20 In comparison, the decaylength (δ; see earlier) of the acoustic wave in the fluid (PBS) fora 62-MHz resonator is ∼67 nm. Since this value is roughly 1 orderof magnitude larger than the protein film thickness, the QCMsurface load from HSA can be treated as a finite viscoelastic layerin a fluid ambient. Bandey and co-workers provided the full one-dimensional model of responses of a thickness-shear moderesonator under these loading conditions, and the surface me-chanical impedance Zs is represented by.21

Zs ) Zsfilm[Zs

fluid Cosh(γhf)+ Zsfilm Sinh(γhf)

Zsfilm Cosh(γhf)+ Zs

fluid Sinh(γhf)] (9)

where Zsfluid is the characteristic mechanical impedance of a

Newtonian fluid given by

Zsfluid)(ωF1η1

2 )1⁄2(1+ j) (10)

In eq 10, Fl and ηl are the liquid density and shear viscosity,respectively. The viscoelastic HSA film is characterized by itscomplex shear modulus according to Gf ) G′ + jG″, where G’and G′′ are the storage and loss modulii, respectively. The surfacemechanical impedance Zs

film can be written as

Z sfilm ) (GfFf)1⁄2 tanh(γhf) (11)

where Ff and hf are the film density and thickness, respectivelyand γ is the complex wave propagation constant given by

γ) jω(Ff ⁄ Gf)1⁄2 (12)

In order to analyze for the viscoelastic response behavior thevalues of Ff and hf were estimated to be ∼1040 kg ·m-3 and ∼ 7nm, based as on the reported HSA aeral density of ∼750 ng · cm-2

at saturation coverage with the thickness set at the lower limit ofthe range 7-10 nm.20 The storage and loss modulii (G′, G′′) wereassumed to be 6.6 × 104 and 7.1 × 105 N ·m-2, respectively, basedon values taken from published work on mussel adhesive proteinonanonpolarsubstrate.22AssumingthevalidityoftheButterworth-VanDyke model for the quartz resonator, we used eq 9 to calculatethe complex mechanical impedance as seen by the quartzresonator. Since the motional capacitance Cm of the quartzresonator remains unchanged during the experiment, the imagi-nary part of the mechanical impedance implies a change in themotional inductance. Therefore, the quartz series resonancefrequency as a function of the protein surface coverage and canbe explicitly written as

∆f) 12π[ 1

√CmLm

- 1

√Cm(Lm + Lm′ )] (13)

where L′m is the motional inductance contributed due to adsorp-tion of the protein layer.

The values of the static and motional capacitances, C0 and Cm,and the motional inductance before and after adsorption of thethiol layer were determined experimentally using the Agilent4294A impedance analyzer. Using the values of C0 ) 4.577 pF,Cm ) 7.437 aF, Lm ) 876.74 mH, and f0 ) 62.325 MHz andapproximating the density and viscosity of the PBS/HSA solutionwith those of water yields a ∆f value of 50.725 kHz. This value isbased on an ideal resonator behaving according to eq 1. If wecorrect for the ratio of the observed and ideal sensitivities, (∆f/∆m)obs/(∆f/∆m)ideal ∼0.2 (see Calibration Experiments), we obtaina frequency shift of 10.145 kHz, closely comparable to theexperimentally measured frequency shift of 11.136 kHz withrespect to air. Furthermore, the frequency shift was not found tobe a very sensitive function of the storage and loss modulii of theprotein layer. For example, changing the value of G′ by a factorof 100 from the reported value for mussel protein22 only resultsin a 6% change in the resonance frequency in our calculations.

CONCLUSIONSWe have reported the design, fabrication and performance of

∼25-µm-thick (62.23 MHz), 500-µm-diameter miniaturized quartzresonators as gravimetric sensors. In comparison to a commercial5-MHz QCM, the miniaturized 62-MHz QCM, in spite of a reduced(1/5) mass sensitivity from the ideal value, shows ∼170 timeslarger frequency change and ∼20 times higher signal-to-noise ratiofor saturation HSA adsorption on a hydrophobic thiol SAMssurface. In the high-concentration limit, the equilibrium adsorptionrate constant for the HSA protein was determined to be K ) 8.03× 106 M-1, which is in reasonable agreement with the valuesreported in the literature. Furthermore, due to the reducedpenetration depth of the acoustic wave in the sampling fluid, thecomplete multilayered viscoelastic model has to be used to explainthe observed frequency shift rather than a simple mass loading(Sauerbrey) model. This should prove to be a very useful tool for(20) Krishnan, A.; Siedlecki, C. A.; Vogler, E. A. Langmuir 2003, 19 (24), 10342–

10352.(21) Bandey, H. L.; Martin, S. J.; Cernosek, R. W.; Hillman, A. R. Anal. Chem.

1999, 71 (11), 2205–2214.(22) Hook, F.; Kasemo, B.; Nylander, T.; Fant, C.; Scott, K.; Elwing, H. Anal.

Chem. 2001, 73 (24), 5796–5804.

Figure 7. Schematic representation of the various layers presenton the quartz resonator during the HSA adsorption experiment andthe approximations used to model their effect on the quartz resonatorfrequency response.

5935Analytical Chemistry, Vol. 80, No. 15, August 1, 2008

probing the viscoelastic properties of nanometer scale adsorbedbiomaterials. One major issue that currently remains unresolvedis the observation of reduced mass sensitivity by a factor of ∼5and the low phase rotation in the sensor resonance characteristics.We believe that the main cause of this is related to the built-instress in the resonators from the packaging processes. Furtherexperiments are currently underway to determine the cause forthe reduced sensitivity.

ACKNOWLEDGMENTThe authors acknowledge partial financial support from the

NSF funded Pennsylvania State University Center for Nanoscale

Science (MRSEC DMR-0080019) and the use of facilities at thePSU Site of the NSF NNIN under Agreement 0335765. Experi-mental assistance from Yi-Hsiu Liu and Dr. Abhijat Goyal in theinitial setup of the measurements and experimental protocol isalso acknowledged. S.T. acknowledges the support of researchfellowship from the Alexander von Humboldt Foundation andWalton Fellowship from the Science Foundation of Ireland.

Received for review March 14, 2008. Accepted May 14,2008.

AC8005395

5936 Analytical Chemistry, Vol. 80, No. 15, August 1, 2008