Quartz crystal microbalances for quantitative biosensing and characterizing protein multilayers

Biosensors & Bioelectronics Vol. 12. No. 7, pp. 567-575, 1997 © 1997 Elsevier Science Limited All rights reserved. Printed in Great Britain

PII: S0956--5663(96)~77-2

ELSEVIER

0956-5663/97/$17.00

Quartz crystal microbalances for quantitative biosensing and characterizing protein multilayers Jan Rickert, Andreas Brecht & Wolfgang G6pel* Institute of Physical and Theoretical Chemistry and 'Center of Interface Analysis and Sensors' University of TUbingen, Auf der Morgenstelle 8, D-72076 Ttibingen, Germany

(Received 13 August 1996; accepted 20 November 1996)

Abstract: The use of quartz crystal microbalances (QCMs) for quantitative

biosensing and characterization of protein muitilayers is demonstrated in three case studies. Monolayers of QCM-based affinity biosensors were investigated first. Layers of a thiol-containing synthetic peptide constituting an epitope of the foot-and-mouse-disease virus were formed on gold electrodes via selfassembly. The binding of specific antibodies to epitope-modified gold electrodes was detected for different concentrations of antibody solutions. Oligolayers were studied in a second set of experiments. Dextran hydrogels were modified by thrombin inhibitors. The QCM response was used in a competitive binding assay to identify inhibitors for thrombin at different concentrations. Multilayers of proteins formed by self-assembly of a biotin-conjugate and streptavidin were investigated next. The QCM frequency response was monitored as a function of layer thickness up to 20 protein layers. A linear frequency decay was observed with increasing thickness. The decay per layer remained constant, thus indicating perfect mass coupling to the substrate. Frequency changes a factor of four higher were obtained in buffer solution as compared to measurements in dry air. This indicates a significant incorporation of water (75% weight) in the protein layers. This water behaves like a solid concerning the shear mode coupling to the substrate. The outlook discusses briefly the need for controlled molecular engineering of overlayers for subsequent QCM analysis, and the importance of an additional multiparameter analysis with other transducer principles and with additional techniques of interface analysis to characterize the mechanical coupling of overlayers as biosensor coatings. A promising trend concerns the use of QCMarrays for screening experiments. 01997 Elsevier Science Limited Keywords: quartz crystal microbalance, biosensors, affinity, protein multilayers

INTRODUCTION Quartz crystal microbalances (QCMs) are suitable transducers for chemical and biochemical sensing *To whom correspondence should be addressd. Tel: + + 49 7071 29-6904 Fax + + 49 7071 29-6910.

in general (Thompson et al., 1991; Schumacher, 1990). Recent advances in overlayer preparation and transducer electronics made possible the sensitive operation in liquids for qualitative studies of a variety of affinity reactions (Suleiman & Guilbault, 1991). The formation of sensitive coatings, the incorporation of test molecules and inter567

Jan Rickert et al.

Biosensors & Bioelectronics

fering effects from, for example, incorporation of water, determine the resonance frequency before and after a specific modification. The quantitative understanding of QCM frequency responses on the molecular level, however, is still missing for this application. It is nevertheless mandatory for calibrations and for systematically optimizing sensitivities, selectivities and stabilities. The QCM frequency response is influenced by a number of interfacial properties, such as the mass, the effective viscosity and stiffness (Noel & Topart, 1994), the conductivity (Yao & Zhou, 1988), the dielectric constant (Shana et al., 1994) or the electrode morphology (Urbakh & Daikhin, 1994) and is hence influenced by a general very complex mechanism of acoustic coupling. The present study aims at a quantitative understanding of QCM responses in bioaffinity measurements, which are monitored under well-controlled experimental conditions using mono-, oligo- and multilayers.

BASIC PRINCIPLES The QCM experimental setup and evaluation scheme based on monitoring impedances as a function of frequency with their absolute values and phase shifts have been published previously (Rickert et al., 1996a). The long-term aim of our measurements is to vary systematically sensitive coatings, test molecules and conditions with different interfering effects. In general the frequency shift Af

Af = ~fmAm + ~ f A71+ ~pAp ~f + ~-T~T +...

(1) can be investigated in view of the different influences from the mass m, the viscosity r/, the density p, the temperature T, etc. In the extreme of perfect coupling, the test molecules are attached rigidly to the surface and therefore influence directly the acoustic shear wave. If other parameters are constant and only mass changes play a role Af = - C~Am/A

(2)

holds with C as a constant, fo as the resonance frequency and A as the surface area. The constant C is a function of quartz crystal parameters like shear modulus, density and geometric parameters (Sauerbrey, 1959). 568

In the other extreme, frequency changes in a homogeneous liquid are just determined by viscosity and density changes and

Af = - Df~o/2(p"r))°"5

(3)

holds with D as a constant. In bioaffinity sensors complex overlayers are formed and are modified during molecular recognition processes. In these applications frequency variations may only in the simplest cases be described by either of the two models alone.

EXPERIMENTAL AND EVALUATION SCHEMES The monolayer and multilayer experiments were performed in a flow-through cell, the details of which are described elsewhere (Rickert et al., 1996a). The affinity assays with thrombin and thrombin inhibitors were performed in a stopped-flow cell which has the same geometry as a common microtiter plate well. AT-cut QCMs (8.5 mm diameter, resonance frequency 10MHz, theoretical value for C = 8, 1 Hz/(ng/mm2) in Eq. (2)) were obtained from Kristallverarbeitung Neckarbischofsheim (Germany). QCMs were mounted on the bottom of cavities drilled in a Perspex TM plate (6mm diameter, 9 m m center-to-center spacing, corresponding to a standard microtiter plate). A miniaturized oscillator electronics was attached underneath. The resonance frequencies were registered with a time resolution of 2 s by a computer-driven data acquisition system.

Kinetics of QCM responses The binding of thrombin to the dextran hydrogel is registered as a function of time by monitoring the resonance frequency of the QCM. The binding kinetics is diffusion controlled and is monitored in a resting solution after addition of analytes. The diffusion rate as a function of time is determined by the second law of Fick. Theoretical diffusion profiles at different times are plotted in Fig. 1. The mass transport to the surface of the sensor is derived from an integration of these curves. The latter is proportional to the square root of time with the corrsponding frequency change of the QCM given by:

Af = fo.A.t 1/2

(4)

Quartz crystal microbalances

Biosensors & Bioelectronics 1 s 10s 60s

300 s

900 s

3

80 -

l

. dr: 5 > GO2 J 0 40$ 20 -

0.8 0.6 0.4

measured data linear regression

0.2 O-

0.0 0.00

0.25

0.50

0.75

0

1.00 x

100

200

300

[mm1

Fig. 1. Relative concentration as a function of the distance from the sugace. Diffusion profiles are plotted for different times. The extrapolated difision coefficient of thrombin was chosen for this presesntation.

A square root fit of the curve was used to quantify the rate of signal change by determing the parameter A. Calibration of QCMs As the Sauerbrey theory is based on the idealized assumptions that mass loading of the QCM shows ideal acoustic coupling to the crystal surface and the crystal is an infinite plane, a calibration of the QCM with protein solutions was performed first. Protein (bovine serum albumin, molecular weight 68 000 g/mol) was soluted in bidestilled water at concentrations of 20-80 mg/l. Then 5 ~1 drops with 100400 ng of protein were deposited on the QCM surface and dried in argon at 295 K until a stable resonance frequency was obtained. Under these conditions all water with the exception of the structural water, which accounts typically only a few percent of the total mass, was evaporated (Bone & Zaba, 1992). To ensure a homogeneous mass loading in these experiments, the QCM was adjusted horizontally, and prior to the protein deposition the gold electrode was prepared to be hydrophilic by a self-assembled monolayer of whydroxyundecanethiol and the surrounding silica was prepared hydrophobic by silanization. The calibration graph (frequency change as a function of deposited protein mass) was linear and a sensitivity of 0.17 + 0.015 Hz/ng (or 6.5 f 0.58 Hz ng-’ mm2 if related to the unit area) was found (see Fig. 2). Every datum point was measured at least five times. These values are 20% lower than predicted from the simple theory of Sauerbrey for an infinite quartz crystal.

400 mass [ng]

Fig. 2. Frequency decrease as a function of deposited mass of protein. The data represent average values of six to 10 single measurements. The standard deviation is indicated by error bars. The linear regression curve is used as a calibration curve, its slope amounts to 0.17 H&g, corresponding to the area-related value of 6.5 H/(ng/mm2).

They are attributed to the finite size of the quartz. Deviations of this order were observed by other authors. As just one representative example, metal atom deposition leads undoubtedly to mass changes only, without any influences of viscosity. In this context Ward and Delawski found a mass sensitivity for copper electrodeposition which was 25% lower than the theoretical value (Ward & Delawski, 1991). Affinity sensors Monolayers:

self-assembled

peptides

Controlled monolayer formation was performed by using the self-assembly technique. In these experiments a synthetic antigen of the foot-andmouth-disease virus (FMDV) was immobilized via thiol-gold coupling to Au( 111) electrodes of the QCMs. The availability of the immobilized antigen for immunological recognition by a specific antibody was demonstrated by an enzymelinked immuno-sorbent assay (ELISA) (Knichel et al., 1995). The same immunological reaction, i.e. the determination of antibodies against the FMDV, was then monitored with the QCM mounted in a flow-through cell. Typical mass changes upon addition of anti-FMDV-antibody solution (50 pg/ml in PBS) were of the order of 38 Hz for 6 MHz quartzes, and 195 Hz for 12 MHz quartzes. Details have been published elsewhere (Rickert et aE., (1996a, b)). Concentration-dependent signals were obtained which were in line with corresponding signals obtained

Jan Rickert et al.

Biosensors & Bioelectronics

independently in measurements of complex impedances (Rickert et al., 1996c; Knichel et al., 1995). The sensitive coating of the QCM could be regenerated for repeated use as a mass-sensitive biosensor elsewhere (Rickert et al., 1996a).

<3 -20

In the next set of experiments we used QCMs for a competitive binding assay with mass-sensitive detection. The preparation of dextran coatings on gold electrodes of quartz microbalances with a thrombin inhibitor as reference substance is shown schematically in Fig. 3. A typical timedependent frequency change in a stopped-flow experimental arrangement is shown in Fig. 4. Curves are fitted as shown in the previous paragraph (Eq. (4)). The fit parameter A characterizes the concentration of unbound thrombin. For low thrombin concentrations a linear dependence is found as shown in Fig. 5. Results obtained in a binding inhibition assay with increasing concentration of a thrombin-inhibitor (0-1/xM thrombin, 0.8-850/xM inhibitor, 600 g/mol molecular weight) are shown in Fig. 6. The corresponding relative fit parameters A/Ao as a function of the inhibitor concentration in the assay are shown in Fig. 7. Here, Ao denotes the fit parameter at zero inhibitor concentration. The oligolayer thickness is on the order of

gold electrode

OH I

OH I

-O;-O

r ~ r fitting I ~

data

fitting function: f=At°~+fo

I ~

-60

additionof test solution I

I

I

I

I

I

I

-2

0

2

4

6

8

10 t [min]

Fig. 4. Evaluation scheme for the stopped-flow binding experiments with determination of the fit parameter A.

20 nm. The affinity sites are distributed statistically in the hydrogel. Reversible adsorption and desorption experiments and reproducible QCM results were obtained. The concentration-dependent signals are in line with corresponding results obtained independently with interferometric optical transducers (Piehler et al., 1996) and lead to comparable shapes in calibration curves. The latter describe the changes in frequency or in optical thickness as a function of inhibition concentration.

H

H2

NH NH C-O C-O I

~tethanolic solutionof hiolcarbonacid

~ - t i m e interval'-~

-40

Oligolayers: modified dextran hydrogels

~

o measured data - - fitted data

I

uplingof inodextran EDC

2

~

couplingof reference inhibitorwith DIC l-carbodiimide)

2

yl-dimethylinopropylbodiirnide)

Fig. 3. Schematic illustration of the preparation of dextran coatings on gold electrodes of QCMs.

570

Biosensors & Bioelectronics "E

Quartz crystal microbalances .----W-

1

100,

¢-

0

"-~,.,

N= 3

oO

75

<

~- 50 09 aJ

E

LL

6_ x -3 LU

25

""".,,...

m

13. -2

"".. N = 4 N = Number of experiments I

0

i

I

J

I

"'"'-~.J2 i

I

I

=

1 10 100 100010000 concentration inhibitor [nM]

I

1 2 3 4 5 C o n c e n t r a t i o n T h r o m b i n [pg/ml]

Fig. 7. Concentration-dependent results from the competitive immuno assay with relative signals A/Ao.IO0% plotted as a function of the test inhibitor concentration.

Fig. 5. Experimental fit parameter A for different concentrations of thrombin.

20 0 "r ,d -40

add,,iool of test so lution I

-80

OnM

-16o' 6 '16o'2bo' 3bo time [S] Fig. 6. Characteristic frequency changes in a binding inhibition assay with a test inhibitor (molecular weight 600 g/mol) at concentrations between 0.8 and 850 nM for a fixed concentration of thrombin (lOOnM = 3.3 txg/ml).

Multilayers: biotin-BSA/SA sandwiches In an attempt to analyze quantitatively mass changes related to frequency changes as a function of layer thickness we investigated ordered multilayers of proteins. We chose a model system which was formed by self-assembly of a biotinconjugate and streptavidin. The QCM frequency response was monitored in situ during layer formation. Protein multilayers were built up by successive incubation with a biotinylated bovine serum albumin (biotin-BSA) and polymerized streptavidin (SA) with a time sequence as illus-

trated schematically in Fig. 8. As very low flow rates were used, their influence on the resonance frequency was proven to be negligible. The binding kinetics were recorded in real time (resolution less than 5 s) with typical results indicated in Fig. 9. Surprisingly, the frequency decrease per layer is independent of the layer thickness up to 20 monolayers. Deviations in the first monolayer are due to the physisorption of protein at the substrate as compared to the much stronger interaction between the biotin conjugate and the streptavidin in all subsequent layers. It is also surprising to note that even the 20-monolayer-thick protein overlayers still behave like a solid without any influence from a gradual transition towards less pronounced acoustic coupling (which might require one to describe the coupling via changes in the viscosity or stiffness). Optical investigations on the same affinity system with spectroscopic ellipsometry indicate an increase of layer thickness of 20 nm per protein layer and a water content of the protein multilayer system of about 70 + 5% (Spaeth et al., 1997). From these optical measurements a layer thickness of 400nm for the 20 monolayers was deduced. These values make it possible to draw conclusions about the acoustical characteristics of this protein layer system. Evidently the layer consists of a highly hydrated, protein meshwork. One likely explanation of the frequency decrease observed during the formation of the protein layer system would be viscosity effects (Eq. (3)). The protein multilayer may be con571

Jan Rickert et al.

Biosensors & Bioelectronics 1L 5968797.0

5968.85

-r ~ ' 5968797.5

T

I

5968.80

r

>

I .

i

i

40

45

50

55

5968.75

i ~iotin-BSA i

5~8.70

0

i

4

30

60 64

90

120

time U'" Jr--;n1

•

Fig. 8. Time dependence of the multilayer preparation of biotin-BSA/SA layers and changes in flow rates to produce reproducible QCM data by an optimized preparation protocol.

T100-<3 50-

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1 9 2 0 layer no.

Fig. 9. Frequency decrease per layer obtained from the on-line registration of the resonance frequency during the preparation of a protein multilayer. Deviations occur for the first monolayer because of its weaker binding to the substrate.

sidered to form a fluid layer with increased viscosity in comparison to the surrounding aqueous phase but poor coupling to the QCM oscillation. The response of the QCM to such viscosity effects decays exponentially with thickness, as for 6 MHz QCMs the penetration depths for the acoustic wave is about 1/xm. For the spectroscopically determined thickness of 400 nm (corresponding to 20 protein layers), a decrease of the frequency response per layer to about 0.7 would be expected if viscosity effects were dominant. Such a decrease in frequency decrease per layer is not observed. We therefore conclude that the entrapped water behaves like a solid concerning the coupling of the transversal shear vibrations of the underlying QCM electrode. 572

Additional experiments support this conclusion. After preparing the layer systems in PBS they were carefully washed with bidistilled water and dried in a dry argon flow, and subsequently the equilibrium changes in the resonant frequencies were determined in air. These changes were significantly smaller than those monitored in buffer. These findings are in line with results from other authors (Caruso et al., 1995; Geddes et al., 1994), who also observed that frequency changes upon protein attachement in water exceed by far frequency changes in air. We determine an increase of the response by a factor of 4.13 as mean value (see Table 1). Deviations were found in the mass changes per layer. They are due to different starting conditions for the preparation of the multilayers and slight differences in the molecular

Biosensors & Bioelectronics

Quartz crystal microbalances

TABLE 1 Comparison of frequency decreases of 6 MHz quartz crystals measured on-line in PBS and calculated from the resonance frequency measured before and after the protein multilayer preparation in air for two different layer thicknesses (a detailed discussion is given in the text) No. of layers Af in PBS Am per mass and area in PBS according to the Sauerbrey equation Af in air Am in air per layer and area calculated from the calibration Ratio (Af in PBS)/(Af in air)

weight of the polymerized SA or the amount of biotin molecules attached to one BSA molecule. Based on the calibration described above we estimate an average mass loading of 5 ng/mm 2 per protein layer in air. As no viscosity effects are observed during the preparation of the multilayer in aqueous buffer, the increase in frequency changes in buffer as compared to air results from entrapped water. Even the absolute value of the protein mass deposited agrees with that obtained from optical measurements. As the density p of hyrated proteins does not differ significantly from the density of water (p = 1 g/cm 3) a 20 nm thick protein layer weighs 20 ng/mm 2. With a water content of 70% the protein mass loading estimated by optical measurements is 6 ng/mm:. From this we conclude that the differences in the QCM sensor signal between the dry protein layer and the wet protein layer are determined by the water content in the latter. Surprisingly, this entrapped water determines 75% of the total mass of the layer.

CONCLUSION For the monolayers investigated in our studies independent mass calibrations were not available but highly desired for further quantitative evaluation of concentrations of epitopes and antibodies. For the oligolayers investigated in our studies the experimental error in independent absolute calibration of masses was relatively high (30%). Nevertheless, the data indicate that viscosity and density changes are also negligible for these layer systems. For the multilayers investigated in our studies, we first determined the calibration factor of QCMs for rigid overlayers in argon. A quantitative determination of mass changes was possible in these multilayers. In the prototype system

10 1342.0 Hz 16.49 ng/mm: 326.6 Hz 5.05 ng/mm2 4.11

4 616.0 Hz 18.91 ng/mm 2 148.0 Hz 5.73 ng/mm: 4-16

chosen for this study, any influence from changes in viscosity or density on frequency changes can be neglected even for 20 monolayers of proteins. The results also show that in certain layer systems water has to be treated like a solid.

OUTLOOK The data presented show the surprising result that for certain model systems up to large thicknesses (here, 400 nm) a simple mass variation explains quantitatively the frequency changes observed with the QCM. On the other hand, liquids or less strongly bound macromolecules like DNA strings on QCMs show more comlex influences of frequency decreases with layer formation (as listed globally Eq. (1), which in the next stage of development could be modified by introducing information about local bonding in inhomogeneous polymeric or supramoleclar layers). Two consequences follow for future work. (1) The use of QCMs for sensitive pharmascreening or for general screening experiments of new substances is suitable. A variety of monolayer and oligolayer systems may thus be investigated. Since mass changes are a general phenomena which occur in any molecular recognition, such QCM sensors will now be used in binding assays for drug screening as illustrated schematically in Fig. 10. Preliminary results will be published elsewhere (Brecht et al., 1996). (2) A prototype of overlayer systems should now be studied with systematically varied properties between the extreme of a liquid coupling with its different parameters, like density and viscosity describing homogeneous bulk parameters (Eq. (3)), and the other extreme of a solid (Eq. (2)). Carefully 573

Jan Rickert et al.

Biosensors & Bioelectronics

/~:~ ~:;~1~ /~/~

~; ~

array of quartz ~-Y crystalmicrobalances

i i

1___ mixing

1.!1 ,

I Computer

c''cui"

with referencefrequency data collectionwith personal computer

Fig. 10. Construction of an array of QCMs for mass-sensitive screening experiments. Quartz crystals with a resonance frequency of 10 MHz are placed in cavities of a microtiter plate. The miniaturized oscillator electronics is attached underneath. The resonance frequencies are mixed with a reference signal and registered in parallel by a computer-driven data aquisition system.

chosen homogeneous model systems of liquid-like and solid-like overlayers as well as their interpretation from QCM results will then make it possible also to understand more complex systems such as DNA and other high molecular weight proteins with the same methods. The careful preparation of such model systems requires a molecular engineering of overlayer formation and independent interface analysis of such structures as discussed earlier (Gtipel, 1995).

ACKNOWLEDGEMENTS The authors thank Tilo Weiss for preparing the gold electrodes of the quartz crystals. They also gratefully acknowledge Boehringer Mannheim (Germany) for the donation of the Biotin-BSA and SA and BASF (Germany) for donation of thrombin and inhibitors. This work was supported 574

by the Bundesministerium fur Forschung und Bildung under project no. 0310838 and by the Fonds der Chemischen Industrie.

REFERENCES Bone, S. & Zaba, B. (1992). Bioelectronics. Johm Wiley, Chichester, pp. 89129. Brecht, A., Burckardt, R., Rickert, J., Stemmler, I., Schlitz, A., Fischer, S., Friedrich, T. & G~pel, W. (1996). Transducer based approaches for parallel binding assays in HTS. J. Biomolecular Screening 1, 191-201. Caruso, F., Serizawa, T., Furlong, D. N. and Okahata, Y. (1995) Quartz crystal microbalance and surface plasmon resonance study of surfactant adsorption onto gold and chromium oxide surfaces. Langmuir 11, 1546-1552. Geddes, N. J., Paschinger, E. M., Furlang, D. N., Ebara, Y., Okahata, Y., Than, K. A. and Edgar, J. A. (1994) Piezoelectric crystal for the detection of immunoreactions on buffer solutions. Sensors & Actuators B 17, 125-131.

Biosensors & Bioelectronics

Gtipel, W. (1995) Interface analysis in biosensor design. Biosensors & Bioelectronics 10, 853. Knichel, M., Heiduschka, P., Beck, W., Jung, G. and G6pel, W. (1995) Utilisation of a self-assembled peptide monolayer for an impedimetric immunosensor. Sensors & Actuators B 28, 85-94. Noel, M. A. M. and Topart, P. A. (1994) High-frequency impedance analysis of quartz crystal microbalances. I. General considerations. Anal. Chem. 66, 484~J~91. Piehler, J., Brecht, A., Geckeler, K. E. and Gauglitz, G. (1996) Surface modification for direct immunoprobes. Biosensors & Bioelectronics 11, 579-590. Rickert, J., Weil3, T. and Gt~pel, W. (1996b) Selfassembled monolayers for chemical sensors: molecular recognition by immobilized supramolecular structures. Sensors & Actuators B 31, 45-50. Rickert, J., WeiB, T., Kraas, W., Jung, G. and G~pel, W. (1996a) A new affinity biosensor: selfassembled thiols as selective monolayer coatings of quartz crystal microbalances. Biosensors & Bioelectronics 11, 591-598. Rickert, J., Gi3pel, W., Beck, W., Jung, G. and Heiduschka, P. (1996c) A 'mixed' self-assembled monolayer for an impedimetric immunosensor. Biosensors & Bioelectronics 11, 757-768. Sauerbrey, G. (1959) Verwendung yon Schwingquarzen zur W~igung dunner Schichten und zur Mikrowagung. Z. Phys. 155, 206--222.

Quartz crystal microbalances

Schumacher, R. (1990) The quartz microbalances. A new measuring rechnique for in situ investigation of solid-liquid phase boundaries. Angew. Chem. Int. Ed. Engl. 29, 329. Shana, Z. A., Zong, H., Josse, F. and Jeutter, D. C. (1994) Analysis of electrical equivalent circuit of quartz crystal resonator loaded with viscous conductive liquids. J. Electroanal. Chem. 379, 21-33. Spaeth, K., Brecht, A. & Gauglitz, G. (1997). Studies on the siolin-avidin multilayer adsorption by spectroscopic ellipsometry. Colloid Interface Sci., (in press). Suleiman, A. A. and Guilbault, G. G. (1991) Piezoelectric immunosensors and their application. Anal. Lett. 24, 1283-1292. Thompson, M., Kipling, A. L., Duncan-Hewitt, W. C., Rajakovic, L. V. and Cavic-Vlasak, B. A. (1991) Thickness-shear-mode acoustic wave sensors in the liquid phase. A review. Analyst 116, 881-890. Urbakh, M. and Daikhin, L. (1994) Influence of the surface morphology on the quartz crystal microbalance response in a fluid. Langmuir 10, 2836-2841. Ward, M. D. and Delawski, E. J. (1991) Radial mass sensitivity of the quartz crystal microbalance in liquid media. Anal. Chem. 63, 886-890. Yao, S. -Z. and Zhou, T. -A. (1988) Dependence of the oscillation frequency of a piezoelectric crystal on the physical parameters of liquids. Anal. Chim. Acta 212, 61-72.

575