A

arstwlp©

K

1

lSutctbbmSdeAcf

0d

Available online at www.sciencedirect.com

Biosensors and Bioelectronics 23 (2008) 833–838

On-chip manipulation and detection of magnetic particlesfor functional biosensors

X.J.A. Janssen a,∗, L.J. van IJzendoorn a, M.W.J. Prins a,b

a Eindhoven University of Technology, Eindhoven, The Netherlandsb Philips Research, Eindhoven, The Netherlands

Received 25 June 2007; received in revised form 23 August 2007; accepted 31 August 2007Available online 6 September 2007

bstract

We demonstrate the real-time on-chip detection and manipulation of single 1 �m superparamagnetic particles in solution, with the aim to developbiosensor that can give information on biological function. Our chip-based sensor consists of micro-fabricated current wires and giant magneto

esistance (GMR) sensors. The current wires serve to apply force on the particles as well as to magnetize the particles for on-chip detection. Theensitivity profile of the sensor was reconstructed by simultaneously measuring the sensor signal and the position of an individual particle crossinghe sensor. A single-dipole model reproduces the measured sensitivity curve for a 1 �m bead. For a 2.8 �m bead the model shows deviations,

hich we attribute to the fact that the particle size becomes comparable to the sensor width. In the range between 1 and 10 particles, we observed a

inear relationship between the number of beads and the sensor signal. The real-time detection and manipulation of individual particles opens theossibility to perform on-chip high-parallel single-particle assays.

2007 Elsevier B.V. All rights reserved.

nction

sbm

wsipitoscfi

eywords: Magnetic biosensor; GMR detection; Superparamagnetic beads; Fu

. Introduction

Biological molecules are commonly detected with the aid ofabels, for example fluorescent dyes or enzymes (Wild, 2005).ince a number of years biosensors are being investigated whichse superparamagnetic particles as labels. One of the advan-ages of this approach is that the magnetic background of evenomplex biological fluids is very low. It is particularly advan-ageous to detect such magnetic labels in a chip-based device,ecause of the integration and miniaturization potential. Chip-ased detection of magnetic labels has been demonstrated usingagneto-resistive sensors (Rife et al., 2003; Graham et al., 2004;hen et al., 2005; Han et al., 2006; Wirix-Speetjens et al., 2006;e Boer et al., 2007; Megens et al., 2007) and Hall sensors (Besset al., 2002; Ejsing et al., 2004, 2005; Fukumoto et al., 2005;

ytur et al., 2006). Biosensing generally aims at detecting theoncentration of specific biological molecules in a fluid. Weoresee that next-generation biosensors are not only able to mea-

∗ Corresponding author. Tel.: +31 402474663; fax: +31 40 247 2598.E-mail address: x.j.a.janssen@tue.nl (X.J.A. Janssen).

btnt

tt

956-5663/$ – see front matter © 2007 Elsevier B.V. All rights reserved.oi:10.1016/j.bios.2007.08.023

al biosensor; Bond force

ure the presence and concentration of biological molecules,ut can also actively probe for functional information of theolecules.Fig. 1(top panel) sketches an example, namely a format in

hich a molecule is sandwiched between a label and a chipurface. Once bound to the chip surface, the target molecules actuated via the attached particle and the movement of thearticle is detected. The combined actuation and detection cann principle give information about the properties of the cap-ured molecule, e.g., the mechanical compliance of the moleculer its binding affinity. Functional properties are already beingtudied in so-called single-molecule experiments, e.g., with opti-al tweezers, magnetic tweezers (Bockelmann, 2004) or atomicorce microscopy (Evans, 2001). These single-molecule exper-ments are performed in research laboratories and cannot easilye applied in practical settings due to the size and delicacy ofhe equipment. Another challenge is that single-molecule assayseed to be performed with high parallelism in order to get sta-

istically meaningful results in a short time.

Our aim is to develop miniaturized chip-based biosensingechnologies that are suited for real-time high-parallel func-ional assays. Panhorst et al. (2005) described a functional assay

834 X.J.A. Janssen et al. / Biosensors and B

Fig. 1. (Top panel) The target molecule is sandwiched between the surfaceand the bead by capture molecules. The amount of bound beads bound is amop

wtoscwlsfoctm

2

2d

aabmowna

n

mdGatbGmtiGhal

fittsscw

ibmsffmagnetic field caused by increasing number of beads or dueto movement of the beads, result in changes in the resistanceof the GMR and therefore in changes of the amplitude of theside-bands.

easure of the concentration of the target. (Bottom panel) The functionalityf the bonds/molecules is probed by applying force to the bead. For exampleulling the bound bead from the surface with a current wire.

ith magnetic particles on a chip. The particles were subjectedo magnetic fields from on-chip current wires and the releasef individual particles was monitored using an optical micro-cope. For miniaturization purposes, we are interested in sensoroncepts that are fully integrated on a chip. In this paper weill describe a sensor device technology using magnetic-particle

abels, on-chip current wires and on-chip magneto-resistive sen-ors (see Fig. 1(bottom panel)). The current wires serve to applyorces on the particles as well as to magnetize the particles forn-chip detection. We will describe the design of the sensorhip and electronic signal processing, and we will demonstratehe on-chip manipulation and detection of single and multiple

agnetic particles.

. Materials and methods

.1. GMR sensor with integrated excitation for particleetection

The Philips biosensor (Prins and Megens, 2007; de Boer etl., 2007) is designed for measuring concentrations of biologicalgents in fluids using superparamagnetic beads as a label. Theiosensor consists of a magnetic field sensor (based on the giantagneto resistance effect), flanked by two current wires on top

f a silicon substrate (Fig. 2). The sensor and wires are coveredith a layer of silicon nitride and a thin gold layer. The silicon

itride insulates the wires from the fluid while the gold providessurface for biochemical experiments.

In absence of an external magnetic field, the superparamag-etic beads do not have a magnetic moment, but once an external

Frw0s

ioelectronics 23 (2008) 833–838

agnetic field is applied, the beads become magnetized. Theipole-field of the beads causes a change in resistance of theMR strip. This change is a measure for the magnetic field

nd therefore a measure for the amount of magnetic beads onhe sensor surface. In the biosensor the external field is appliedy running a current through (one of) the wires (Fig. 2). TheMR sensor is only sensitive for the in-plane component of theagnetic field so ideally the field of the wires remains unde-

ected while only the horizontal component of the dipole-fields detected. However the centers of the current wires and theMR sensor are not in the same plane, resulting in a smallorizontal field component at the position of the sensor. Sosmall signal is measured even in the absence of magnetic

abels.The 1/f noise is reduced by applying an alternating magnetic

eld to the beads by running an ac current Ie at frequency fehrough the excitation wires. The change of the resistance ofhe GMR at this frequency fe is detected, which increases theignal/noise ratio. Due to the parasitic capacitance between theensor and the wire there is a cross-talk that contributes signifi-antly to the measured signal. Therefore we run an ac current Isith frequency fs through the GMR sensor, with fs � feThe contribution in the output-signal, caused by the change

n resistance at frequency fe (due to the magnetic field of theeads Hbead) is modulated in amplitude with frequency fs. Thisodulated signal appears in the frequency spectrum as two

ide-band peaks, one at frequency fe − fs and one at frequencye + fs. The cross-talk is not amplitude modulated and is there-ore visible as a single peak at frequency fe. Changes in the

ig. 2. Philips biosensor consists of a silicon substrate with a giant-magneto-esistance sensor strip (3 �m × 100 �m × 0.04 �m) flanked by two conductingires (3 �m × 100�m × 0.35 �m). The sensor and wires are covered with.5 �m of silicon nitride and a gold layer. Top figure shows top view of theensor and the bottom figure the cross-section.

and Bioelectronics 23 (2008) 833–838 835

2

d(tbrcotfatmohsai

3

3

ipwtiwiwcaiomwcf

omTw2btsvTtpsp

Fig. 3. Simultaneously measured bead position (upper part) and sensor signal(Tse

ttFcm

((b

3

The magnetic field H at a position r around a current densityj(r) is given by integrating Biot–Savart law:

H(r) =∫

d3r′ j(r′) × (r − r′)4π|r − r′|3

Fig. 4. Sensitivity profile of the sensor measured with one bead crossing thesensor. The position of the left wire is from 0 to 3 �m, the sensor from 6 to 9 �m

X.J.A. Janssen et al. / Biosensors

.2. Experimental setup

The sensor cartridge consists of a molded-interconnecting-evice (Nellissen et al., 2005) with the silicon chip1.4 mm × 1.4 mm in the middle of the MID). The chip con-ains four sensors of which one is covered with SU8 to preventeads from reaching the surface. This sensor can be used as aeference. The cartridge is connected via flex-cable to a printed-ircuit-board that contains processing electronics. The sensorutput is analyzed with a spectrum analyzer (Agilent 4395A)hat measures the amplitude of the two side bands at frequenciese ± fs Hz. The spectrum analyzer is connected via a GPIB-bus tocomputer that is used for recording the measured data. Simul-

aneously with the sensor signal, images are recorded using aicroscope (Leica DM6000M) with a long distance immersion

bjective (Leica HXC APO L63X0.90 W U-V-I) fitted with aigh-speed camera (Redlake MotionPro HS-3). The pictures areaved and during post processing, the positions of the beadsre determined by a 2D cross-correlation of a template and themage.

. Results and discussion

.1. Single bead sensitivity

To study the feasibility of single bead detection with thentegrated current wires and sensor, 3 �l of 1:8000 diluted sus-ension of 1 �m beads (Dynal MyOne) was put on the sensorith a pipet (Transferpette 2–20 �l Brand). The beads started

o sediment due to gravity and after approximately 1 min, de-onized water was added gently to the droplet to make contactith the immersion objective. Brownian motion brought a bead

n the vicinity of the wire. The bead was then attracted to oneire by running a current through it. By alternating switching the

urrent every 10 s between the two wires, the bead traveled backnd forth between the wires. Microscope images were made dur-ng the crossings at a rate of 100 Hz. In each frame, the positionf the bead was determined using the image analysis routine asentioned before. The position of the bead moving between theires is plotted as a function of time (upper part of Fig. 3). The

loser the bead approaches the live wire the larger the attractiveorce becomes so speed of the bead increases.

The sensor signal was measured simultaneously (bottom partf Fig. 3). The measured sensor signal is a superposition of theagnetic field of the bead and that of the wires (Section 2.1).he contribution of both wires is not equal causing the squareave behavior of the sensor signal (Fig. 3) with a periodicity of0 s. This difference in contribution of about 1% can be causedy small variations in the current through the wires or by the facthat the sensor might not exactly centered between the wires. Theignal caused by the magnetized bead crossing the sensor wasisible as a sudden rise in signal followed by a drop in signal.hese spikes clearly correlate with the moment the bead crosses

he sensor strip (dotted vertical lines in Fig. 3). The sensor signallotted versus bead position (open circles in Fig. 4) gives theensitivity profile of the sensor for a 1 �m bead. The sensitivityrofile is measured with the bead traveling from the right wire

afTto

bottom part) as function of time for a single bead crossing the sensor 10 times.he dotted vertical lines indicate the moment the bead crosses the center of theensor. The line connecting the data points in the bottom graph is a guide to theye.

owards the left wire (upper part of Fig. 4) and with the beadraveling from the left wire towards the right wire (lower part ofig. 4). Note that apart from the offset, caused by the non-equalontribution of the wires, the sensitivity profile is the same butirrored.The sensitivity profile was also measured with a 2.8 �m bead

Dynal M-280) in the same way as described above. The resultsclosed circles in Fig. 4) show the same trend as that of the 1 �mead, but the signal changes were higher.

.2. Simulation

nd the right wire from 12 to 15 �m. The open circles give the sensitivity profileor a 1 �m bead, where as the closed circles give the profile for a 2.8 �m bead.he graph shows the data of thirty consecutive crossings. In the top two graphs,

he bead travels from the right to the left and in the bottom two graphs in thepposite direction.

8 and B

ilc

zcf

he

M

wifisb

m

wm

H

wwt

oC

u

wttgco

u

dfaatTw

do

mbwtttmrEtcb2stofTiTmttnsfios

bpowi1ttWmotdto

3

fwtra

36 X.J.A. Janssen et al. / Biosensors

For a current density jy running through a rectangular shapednfinitely long, wire orientated parallel to the y-axis with lowereft corner (x1, z1) and upper right corner (x2,z2), the x-omponent of the magnetic field at the origin is given by

2π

jy

Hx(0, 0) = −[[

z arctanx

z= x

2ln(x2 + z2)

]x2

x1

]z2

z1

Linear coordinate transformation x1 → x1 − x, x2 → x2 − x,1 → z1 − z and z2 → z2 − z gives the horizontal magnetic fieldomponent Hx at any position (x; z). Rotating the coordinaterame gives the vertical field component Hz.

When the superparamagnetic bead is approximated by aomogenous sphere, the magnetization M is parallel with thexternal magnetic field H0 and for low fields given by

= χH0

ith χ the magnetic susceptibility of the bead material includ-ng demagnetization effects. The bead creates a magnetic strayeld that can be approximated by the field of a magnetic dipoleituated at the center of the bead. The dipole moment m is giveny

= MV

ith V the volume of the bead. In coordinate free form theagnetic dipole field is given by

dip(r) = 3(m · r)r − m4πr3

ith r the vector from the position of the dipole to the positionhere the field is being measured, r the absolute value of r, r

he unit vector of r and m the induced magnetic dipole moment.The GMR sensor is only sensitive for the in-plane component

f the magnetic field perpendicular to the long axis of the sensor.onsequently the sensor signal is given by

sensor = SI

wL

∫dx′ dy′ Hdipx (x′, y′)

ith S the sensitivity of the GMR-sensor (about 2.7 �/(kA m)), Ihe GMR sense current of 4 mA, w the width and L the length ofhe sensor strip (Megens et al., 2007). Because the electronicsive an amplification of both the sensor signal and the offsetaused by the magnetic field of the wires, the model is fittednto the data using:

sensor = A + B

∫dx′ dy′ Hdipx (x′, y′)

The distance between the sensor surface and the simulatedipole is in the simulations equal to the radius of the beads (0.5 mor the 1 �m bead and 1.4 �m for the 2.8 �m bead) simulatingbead rolling/skidding over the surface. For the 1 �m bead the

mplitude B is 3 �V and the offset A is 41.3 and 46.5 �V. Forhe 2.8 �m bead B is 27 �V and an offset A of 41.5 and 47.8 �V.he data for both the 1 and 2.8 �m beads are shown in Fig. 4

here the lines represent the model fit.For the 1 �m bead the shape of the measured curve is repro-

uced by the simulation. However for the 2.8 �m bead the shapef the measured and simulated curves are not the same. The

bn2u

ioelectronics 23 (2008) 833–838

easured profile is shifted away from the live wire. This effectecomes more and more visible when the bead approaches theire. Because the size of the bead is comparable to the width of

he sensor, a part of the bead is considerably closer to the sensorhan the center of the bead is. Therefore the bead is detected fur-her away from the sensor than is expected for the single-dipole

odel. The accuracy of the single-dipole model describing theeal situation becomes less with increasing size of the bead.specially when the size of the bead is comparable to, or bigger

han the width of the sensor. The overall shape of the sensitivityurve is comparable with the results found by scanning a 2.8 �mead, attached to an AFM tip, across the sensor (Megens et al.,007). It must be noted that in our experiments, a free bead inolution is detected which is attracted by the live wire. The facthat the bead is free in solution gives rise to Brownian motionf the bead and the crossing time is determined by the magneticorce and the hydrodynamic drag of the bead in the solution.he free bead, also has a variable height above the surface dur-

ng the crossing influencing the shape of the sensitivity profile.his effect is more important for smaller beads because these canove further from the surface due to Brownian motion. Because

he bead traveled close to the surface of the sensor the distanceo the surface depends on the speed of the bead due to hydrody-amic liftoff (Patankar et al., 2001). The change in the simulatedensitivity profile between a bead on the surface and a bead on axed height of 0.5 �m above the surface falls within the spreadf the measurements. Therefore the height of the bead above theensor surface cannot be determined from the measurements.

The magnetic content of the 1 and 2.8 �m Dynal beads haseen determined by Fonnum et al. (2005) using inductively cou-led plasma atomic emission spectrometry (ICP-AES). The ratiof the amount of magnetic material in the 2.8 and 1 �m beadsas found to be 8. In our experiments we measure an increase

n signal amplitude of approximately six between the 2.8 and�m beads (see Fig. 4). We measure a smaller ratio because

he sensitivity decreases with distance to the sensor surface andhe center of the larger bead is located further above the surface.

hen we fit the measured profile in Fig. 4 with the single-dipoleodel we find an increase of the point-dipole strength of a factor

f nine for the 2.8 �m beads compared to the 1 �m beads. Notehat this model overestimates the amount of magnetic materialue to the non-linear decrease in sensitivity with distance fromhe sensor surface and thus provides an upper limit to the amountf magnetic material in the bead.

.3. Bead response curve and bead to bead variations

A linear bead response curve (de Boer et al., 2007) was shownor various numbers of 300 nm beads (50–400). From this itas concluded that the signal change is linearly dependent on

he bead surface density (de Boer et al., 2007). Because of theelative large number of beads, no information can be obtainedbout bead-to-bead variations of the induced magnetic moment

ecause small variations will average out. In this research theumber of 1 �m beads is kept low (compared to de Boer et al.,007) and increased from 1 up to 10 in a single experiment,sing a droplet of 3 �l of 1:6400 diluted 1 �m bead suspension

and Bioelectronics 23 (2008) 833–838 837

(t2Ibb(oaa(toav

acetabnwow4rvptwbvdt

Fbbs

Fm

p1lbχ

4

tcos

X.J.A. Janssen et al. / Biosensors

described in Section 3.1). While the bead is captured abovehe right current wire (20 mA dc), the left current wire (40 mA00 kHz) is used to measure the sensor signal (left part of Fig. 5).t has been calculated that a bead on the right wire is not detectedecause of its relatively long distance to the left wire. Next theead is transported to the left wire by switching off the dc currentmiddle part of Fig. 5). Finally the currents are restored to theirriginal settings (right part of Fig. 5) and the sensor signal isgain measured. The bead is now positioned on the wire with thec current and therefore detected. The decrease in sensor signal�S) (with the bead on the ac wire compared to the situation withhe bead on the dc wire) shows the sensitivity for the detectionf a single 1 �m bead. Note that the change in signal is notffected by the absolute value of the sensor signal which canary between measurements.

After the first measurement cycle, the ac wire is switched offnd the bead is again collected on the dc wire. A bead responseurve has been measured by repeating this measurement cycleach time an extra bead is collected. Averaging the signal beforehe crossing gives the signal with no detected beads while aver-ging the signal after the crossing, gives the signal with detectedeads. The difference in signal called �S is plotted against theumber of beads (Fig. 6). For a low number of beads on theire the change in signal is linear dependent on the numberf beads on the wire. Experiments of de Boer et al. (2007), inhich a number of beads (ranging approximately from 40 to00) was distributed on the surface, show also a linear beadesponse curve. Due to the large number of beads, bead-to-beadariations average out as well as the fact that the sensitivityrofile is strongly position dependent. We got around this posi-ion dependent sensitivity by collecting the beads on a wirehich has a fixed distance to the sensor. Due to the low num-

er of beads on the wire, the effect of averaging bead-to-beadariations is reduced. Some measurements have been done inuplo/triplo (Data points in Fig. 6). Of all possible combina-ions with a difference of only one bead, the change in signal

ig. 5. Comparison of a two measurements to determine the loss in signal peread for a 2.8 and 1 �m bead. The open circles represent the data for the 1 �mead while the closed circles represent the data for the 2.8 �m bead. The cross-ections indicate the position of the bead with respect to the sensor.

fcswsa

fls2ttb

mcttlafmsc

ig. 6. The change in signal for increasing number of 1 �m beads. For eacheasurement, the signal is averaged 180 s before and 180 s after the crossing.

er extra bead is calculated. The mean change in signal per�m bead is 0.081 ± 0.017 �V. This signal variation is much

arger than the noise in the measured signal and is most proba-ly due to bead-to-bead variations in the magnetic susceptibility.

. Conclusion

We have shown the manipulation and detection of the posi-ion of magnetic particles on the surface of a chip with integratedurrent wires and magneto-resistive sensors. Single-particle res-lution is proven for 1 �m particles traveling across the sensorurface. The sensitivity curve of the sensor was reconstructedrom multiple measurements on the same particle. Sensitivityurves simulated with a single-dipole model showed good corre-pondence to curves measured with 1 �m particles. Experimentsith 2.8 �m particles showed deviations with respect to the

ingle-dipole model, probably because the size of the particlespproaches the width of the magneto-resistive sensor strip.

A particle–response curve was measured with 1 �m particles,or particle numbers ranging from 1 to 10. The curve shows ainear relationship between the number of particles and the mea-ured signal. The signal per particle shows variations of about1%, which we attribute to variations in the susceptibility ofhe individual particles. These variations are remarkably smallerhan the signal variations of 70% observed for 2.8 �m particlesy Rife et al. (2003).

In the described sensor chip we can apply forces to theagnetic particles using on-chip current wires. The electronic

ontrol of currents gives a large degree of freedom in terms ofhe size of the forces (from fN to the order of pN forces) andhe time-profile of the forces (�s pulses can be applied, andoading rates of 106 pN/s), both in relevant ranges for biologicalssays. The manipulation and detection of magnetic particles is

ully integrated on the chip and electronically controlled, whichakes the system very compact. The magneto-resistive sen-

ors are deposited on a passive silicon substrate, with electronicircuitry placed external to the chip, which makes the sys-

8 and B

trtmtaram

A

hR

R

A

B

Bd

E

E

EF

F

G

H

M

N

P

P

P

RL.J., 2003. Sens. Actuators A: Phys. 107 (3), 209–218.

38 X.J.A. Janssen et al. / Biosensors

em flexible and cost-effective with respect to technologies thatequire on-chip CMOS electronics (e.g., Hall sensors). Finally,he sensor architecture allows the manipulation and detection of

any individual magnetic particles in parallel, which is impor-ant in order to get statistically meaningful results in a biologicalssay. We envisage that the combination of manipulation andeal-time detection of single beads opens the possibility forssays in which functional information is derived of biologicalolecules.

cknowledgments

The authors thank the project members and students at Eind-oven University of Technology and members from Philipsesearch Eindhoven for their support in this research.

eferences

ytur, T., Foley, J., Anwar, M., Boser, B., Harris, E., Beatty, P.R., 2006. J.Immunol. Methods 314 (1/2), 21–29.

esse, P.A., Boero, G., Demierre, M., Pott, V., Popovic, R., 2002. Appl. Phys.Lett. 80 (22), 4199–4201.

ockelmann, U., 2004. Curr. Opin. Struct. Biol. 14 (3), 368–373.e Boer, B.M., Kahlman, J.A.H.M., Jansen, T.P.G.H., Duric, H., Veen, J., 2007.

Biosens. Bioelectron. 22 (9/10), 2366–2370.jsing, L., Hansen, M.F., Menon, A.K., Ferreira, H.A., Graham, D.L., Freitas,

P.P., 2004. Appl. Phys. Lett. 84 (23).

SWW

ioelectronics 23 (2008) 833–838

jsing, L., Hansen, M.F., Menon, A.K., Ferreira, H.A., Graham, D.L., Freitas,P.P., 2005. J. Magn. Magn. Mater. 293 (1), 677–684.

vans, E., 2001. Annu. Rev. Biophys. Biomol. Struct. 30, 105–128.onnum, G., Johansson, C., Molteberg, A., Morup, S., Aksnes, E., 2005. J.

Magn. Magn. Mater. 293 (1), 41–47.ukumoto, H., Takeguchi, K., Nomura, M., Endo, H., 2005. Solid State Sens.,

Actuators Microsyst. 2, 1780–1783.raham, D.L., Ferreira, H.A., Freitas, P.P., 2004. Trends Biotechnol. 22 (9),

455–462.an, S.J., Xu, L., Yu, H., Wilson, R.J., White, R.L., Pourmand, N., Wang, S.X.,

2006. IEEE Int. Electron Devices Meeting.egens, M., de Theije, F., de Boer, B., van Gaal, F., 2007. J. Appl. Phys. 102,

014507.ellissen, T., Weekamp, W., van Delft, J., Ansems, W., Janssen, E., Prins, M.,

Megens, M., Wimberger-Friedl, R., van Iersel, B., 2005. Proceedings 15thEuropean Microelectronics and Packaging Conference and Exhibition.

anhorst, M., Kamp, P.B., Reiss, G., Brckl, H., 2005. Biosens. Bioelectron. 20(8), 1685–1689.

atankar, N.A., Huang, P.Y., Ko, T., Joseph, D.D., 2001. J. Fluid Mech. 438,67–100.

rins, M.W.J., Megens, M., 2007. Magneto-resistive biosensors. In: Buschow,K.H.J., Cahn, R.W., Flemings, M.C., Ilschner, B., Kramer, E.J., Mahajan,S., Veyssiere, P. (Eds.), Encyclopedia of Materials: Science and Technology.Elsevier.

ife, J.C., Miller, M.M., Sheehan, P.E., Tamanaha, C.R., Tondra, M., Whitman,

hen, W., Liu, X., Mazumdar, D., Xiao, G., 2005. Appl. Phys. Lett., 86.ild, D., 2005. The Immunoassay Handbook, 3rd ed. Elsevier.irix-Speetjens, R., Fyen, W., De Boeck, J., Borghs, G., 2006. J. Appl. Phys.

99, 103903-103903-4.