An antenna-coupled split-ring resonator for biosensing

An antenna-coupled split-ring resonator for biosensingH. Torun, F. Cagri Top, G. Dundar, and A. D. Yalcinkaya Citation: Journal of Applied Physics 116, 124701 (2014); doi: 10.1063/1.4896261 View online: http://dx.doi.org/10.1063/1.4896261 View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/116/12?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Compacted tunable split-ring resonators Appl. Phys. Lett. 103, 162602 (2013); 10.1063/1.4826255 Asymmetric split-ring resonator-based biosensor for detection of label-free stress biomarkers Appl. Phys. Lett. 103, 053702 (2013); 10.1063/1.4816440 Tuning the nonlinear response of coupled split-ring resonators Appl. Phys. Lett. 100, 081111 (2012); 10.1063/1.3689775 Biosensing using split-ring resonators at microwave regime Appl. Phys. Lett. 92, 254103 (2008); 10.1063/1.2946656 Spatially resolved biosensing with a molded plasmonic crystal Appl. Phys. Lett. 90, 203113 (2007); 10.1063/1.2740591

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:193.140.194.112 On: Mon, 29 Sep 2014 20:53:26

An antenna-coupled split-ring resonator for biosensing

H. Torun,1,2 F. Cagri Top,1 G. Dundar,1 and A. D. Yalcinkaya1,2

1Department of Electrical and Electronics Engineering, Bogazici University, Bebek, 34342 Istanbul, Turkey2Center for Life Sciences and Technologies, Bogazici University, Kandilli, 34684 Istanbul, Turkey

(Received 29 May 2014; accepted 11 September 2014; published online 26 September 2014)

An antenna-coupled split-ring resonator-based microwave sensor is introduced for biosensing appli-cations. The sensor comprises a metallic ring with a slit and integrated monopole antennas on top ofa dielectric substrate. The backside of the substrate is attached to a metallic plate. Integrated antennasare used to excite the device and measure its electromagnetic characteristics. The resonant frequencyof the device is measured as 2.12 GHz. The characteristics of the device with dielectric loading at dif-ferent locations across its surface are obtained experimentally. The results indicate that dielectricloading reduces the resonant frequency of the device, which is in good agreement with simulations.The shift in resonant frequency is employed as the sensor output for biomolecular experiments. Thedevice is demonstrated as a resonant biomolecular sensor where the interactions between heparin andfibroblast growth factor 2 are probed. The sensitivity of the device is obtained as 3.7 MHz/(lg/ml)with respect to changes in concentration of heparin. VC 2014 AIP Publishing LLC.[http://dx.doi.org/10.1063/1.4896261]

I. INTRODUCTION

Split-ring resonator (SRR) structures have been usedwidely for various applications in a spectrum altering frommicrowave to photonic bands. Magnetic permeability of SRRstructures becomes negative at resonance when they are excitedappropriately.1,2 This phenomenon has been exploited to engi-neer devices with exotic properties such as negative index ofrefraction.3–5 SRR structures define electrically an LC resona-tor whose resonant frequency is determined by the geometry ofthe structure. SRR structures in centimeter to millimeter-scaleare usually used for microwave band applications.6 Structuresin micrometer-scale are useful for terahertz applications.7,8

Smaller structures in nanometer-scale are introduced for appli-cations in infrared9 and visible light spectrum.10,11

Unlike other types of passive resonator tanks, an SRRstructure typically exhibits sharp resonant behavior withquality factors above 1000 at microwave frequencies.12

Thus, the change in resonant frequency of an SRR structurecan effectively be used as a sensing mechanism. Strain sen-sors have been demonstrated based on mapping the changein geometry of an SRR structure to the change in its resonantfrequency.13,14 Besides, the change in resonant frequencycan also be introduced by a change in dielectric properties ofthe medium. Effective sensors for microfluidic applicationshave been introduced to exploit this phenomenon.15,16 Thegeometry of the structure does not change in this mode ofoperation. The change in dielectric properties of the mediummodulates the effective capacitance of the SRR structure,yielding in the resonant frequency shift. Biosensors havebeen demonstrated using this sensing mechanism for whichbinding of biomolecules on top of an SRR structure altersthe apparent dielectric constant of the device capacitance.Detection of hormones and antigens has been achieved usingan SRR placed in the vicinity of a microstrip line.17 A simi-lar structure was used to measure biotin-streptavidin interac-tions18 and DNA hybridization.19

In this paper, we present an SRR-based biosensor oper-ating at microwave frequencies. Section II presents the struc-ture and operation of devices. The response of the devices toliquid loading through modeling and experimental character-ization is discussed in Sec. III. In Sec. IV, we demonstratethe applicability of the devices for biomolecular detection.Concluding remarks are supplied in Sec. V.

II. STRUCTURE AND OPERATION OF DEVICES

A three-dimensional drawing of the device is shown inFig. 1. A metallic ring with a slit is defined on a dielectricsubstrate to form an SRR structure. The ring is placedbetween two monopole antennas defined on the same sub-strate. The antennas are 29 mm in length, 3 mm in width andare separated by 27 mm. Geometry of the monopole antennasis optimized to perform effectively at 2.5 GHz. SubMiniatureversion A (SMA) connectors for the antennas form the elec-trical ports for the device. The substrate is mounted on top ofan aluminum back plate that provides a ground plane.

We fabricated the current devices on top of a 1.5 mm-thick FR4 substrate using standard printed circuit board

FIG. 1. Three-dimensional drawing of an SRR-based biosensor structureand integrated monopole antennas on a dielectric substrate. The substrate isplaced on top of an aluminum back plate.

0021-8979/2014/116(12)/124701/6/$30.00 VC 2014 AIP Publishing LLC116, 124701-1

JOURNAL OF APPLIED PHYSICS 116, 124701 (2014)

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:193.140.194.112 On: Mon, 29 Sep 2014 20:53:26

manufacturing techniques. The thickness of the metallicstructures is 35 lm. After the definition of metallic struc-tures, we coated the surface of the device with a 10 lm-thickparylene layer. Parylene is a biocompatible polymer that weuse to seal the metallic structures for in-liquid operation. Inaddition, parylene film is used to anchor biomolecules on topof the SRR structure. Parylene films are deposited at roomtemperature using a chemical vapor deposition (CVD) pro-cess and can be patterned at room temperature using oxygenplasma. Parylene layer is hydrophobic in nature and istreated in a plasma chamber using oxygen plasma to allowmolecules to anchor on the surface.20 The purpose of thetreatment is to make the surface rougher rather than etchingthe layer to make it thinner. The radius and the width of theSRR structure are 8 mm and 1 mm, respectively. The widthof the slit is 5 mm. This structure defines electrically an LCresonator that resonates at its magnetic resonance frequency(fm) when it is exposed to a magnetic field perpendicular toits surface. This supports a circulating current along the ring.The structure can be modeled using lumped elements to pre-dict its resonance frequency as expressed in Eq. (1)21

fm ¼1

2pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiCeffLeff

p ; Ceff ¼ Cg þ Cs; (1)

where the effective capacitance (Ceff) and the effective in-ductance (Leff) depend on the geometry of the ring. Theeffective capacitance can be decomposed into two capaci-tance terms parallel to each other. Gap capacitance (Cg)models the capacitance associated with the slit along the cur-rent path.21

Cg ¼ eeffhw

gþ eeff hþ gþ wð Þ; (2)

where h, w and g denote height, width, and slit gap geome-tries of the metallic split ring. Capacitance associated withthe surface charges of a split ring with a radius of r isexpressed by the surface capacitance (Cs) parameter21

Cs ¼ 2eef fhþ wð Þ

pln

4r

g

" #: (3)

As can be seen from Eqs. (2) and (3), both gap and surfacecapacitance terms depend on the effective permittivity (eeff )of the media surrounding the metallic ring. Effective permit-tivity for the device is determined by the FR4 substrate, theparylene layer on top of the rings and additional materialsuch as biomolecules on top of the parylene layer. A changein relative permittivity results in a change in resonant fre-quency of the structure. The calculated resonant frequencyof the SRR structure in air is 2.16 GHz.

A pair of identical antennas is used to excite the deviceand to measure its electromagnetic characteristics. Emittedelectromagnetic waves reflect off of the large aluminumback plate placed within the vicinity of the emitter antenna.Emitted energy couples effectively to the SRR at its reso-nance frequency, fm. This reduces the reflectance measuredat the emitted antenna. In addition, the transmittance meas-ured from the transmitting antenna to the receiving antenna

increases. We measured spectra of reflectance and transmit-tance for sensing applications. The characteristics of theobtained spectra allow us to evaluate the change in relativepermittivity around the SRR structure.

III. EXPERIMENTAL CHARACTERIZATION ANDMODELING

We characterized scattering parameters of the SRR byconnecting the device to a vector network analyzer (ZVB4,Rohde & Schwarz, Munich, Germany) through the ports. Wemeasured reflectance (s11) and transmittance (s21) of the de-vice using the ports shown in Fig. 1. We repeated the meas-urements after we placed a droplet of deionized (DI) water atdifferent locations on the parylene layer covering the metal-lic ring. The volume of each droplet was measured to be16 ll with a precision of 2% using a micropipette. The loca-tions of the droplets are shown in Fig. 2(a). Droplets of DIwater increase the effective permittivity of the SRR struc-ture. Droplets increase either surface or gap capacitance orboth based on the location of the droplets.

Spectra of s11 and s21 are shown in Fig. 2(b). The casewith no droplet is labeled as nominal and the measured reso-nant frequency is 2.12 GHz for this device. Reflectancemeasured at port 1 is high due to the back plate until the

FIG. 2. a) Deionized water droplets with a volume of 16 ll were placed ondifferent locations on the metallic ring as indicated for the experimentalcharacterization. (b) The measured (b) s11 and (c) s21 spectra of the device.

124701-2 Torun et al. J. Appl. Phys. 116, 124701 (2014)

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:193.140.194.112 On: Mon, 29 Sep 2014 20:53:26

resonant frequency of the SRR structure. At this point, s11

decreases sharply. Transmittance between port 1 and 2 islow except the band where SRR resonates. The analyticalmodel of Eqs. (1)–(3) predicts the resonant frequency of thedevice within an accuracy of 2%. The equations do notinclude the coupling effect between the antennas and thering. Apparently, this effect is negligible for the designed ge-ometry as the antennas are placed far away from the ring.

The resonant frequency of the device decreases by plac-ing droplets of DI water due to increased effective capaci-tance. We placed one droplet at a time and dried the surfaceafter gathering data. We observed that the spectra of both s11

and s21 were the same as the nominal one once the surfacewas dry. Two pairs of locations are marked in Fig. 2(a).Locations 1 and 2 are along the symmetry axis of the deviceand correspond to the smallest changes in resonant frequencywhen loaded by DI water. The change in resonant frequencyis 1% for both locations. Location 1 is on the gap of the SRRand a droplet at this location alters the gap capacitance aswell as surface capacitance. The effective capacitance of thecurrent design is dominated by the surface capacitance.Consequently, the effect of locations 1 and 2 is almost thesame. Larger shift in resonant frequency occurs when drop-lets are placed on locations 3 and 4. This corresponds to achange of 14.5% in the resonant frequency. The magnitudeof reflectance and transmittance are different for locations 3and 4. We observed that switching the ports resulted in recip-rocal behavior. This is attributed to the integrated antennasthat excite the device. We also observed that the incidentwave is not a plane wave based on simulations. Propagationpattern of the antenna and the polarization of the incidentwave affect the resonance behavior of the device.

We simulated the scattering parameters of the SRR-based sensor using commercially available electromagneticsimulation software (CST Studio Suite, Darmstadt,Germany). The computational domain includes the SRRstructure, integrated antennas, FR4 substrate, and the metal-lic back plate. We simulated the behavior of the devicesusing ports placed at the bottom of the antennas. The resultsof the simulations are shown in Fig. 3. The resonant fre-quency of the unloaded device was obtained as 2.118 GHz.This is in excellent agreement with the measured frequency.We ran the simulator by placing a semi-sphere to model DIwater on different locations marked in Fig. 2(a). We kept thevolume of the semi-sphere as 16 ll. The s11 and s21 spectraof the device after loading are shown in Figs. 3(a) and 3(b).The resonant frequency of the device matches well with theexperimental results. The mismatch for each of the locationsis less than 1%. The variation of amplitudes at resonance fors11 and s21 are larger. Loss parameters for the metallic struc-tures and the dielectric substrate determine the amplituderesponse of the device. The variation in amplitude can beexplained due to possible mismatches between the assumedand actual loss parameters.

We simulated surface current density across the SRRstructure to verify that the devices undergo magnetic reso-nance at the measured and simulated resonant frequencies.Fig. 4(a) shows the profile of surface current density alongthe metallic ring without any droplet of DI water. Surface

current vectors are distributed evenly with respect to thesymmetry axis of the device at its resonant frequency of 2.12GHz. The device supports a circulating current along the me-tallic ring as expected. The effective magnetic field is per-pendicular to the device at this frequency. Electric fieldvectors are shown in Fig 4(b).

We observed the profile of surface current and electricfield vectors for the cases when a droplet of DI water loadedthe device. Fig. 4(c) shows the profile of surface current den-sity after the 3rd location (see Fig. 2(a)) is loaded. The sym-metry is broken as expected and the current density is highernear the droplet. Nevertheless, the perpendicular magneticfield to the surface of the device supports a circulating cur-rent along the metallic ring. This indicates magnetic reso-nance for the loaded cases as well. Fig. 4(d) shows theelectric field vectors at the resonant frequency. The field vec-tors are rearranged across the gap with a significant decreasein magnitude over the droplet.

Reflectance of the device measured and simulated at theplane of antennas indicates a drop in amplitude at resonance.On the other hand, transmittance peaks up. This is due to theconfiguration of the device including an SRR structure withintegrated antennas on top of a metallic back plate.Scattering parameters of the characterized resonator showthe feasibility of building an oscillator circuit for the readoutof the device. The quality factor of the resonator was meas-ured to be 93 at 2.12 GHz for the unloaded case. The valueof the quality factor varies within 10% for the loaded cases.Maintaining a relatively large quality factor allows

FIG. 3. Simulated (a) s11 and (b) s21 spectra of the device. See Fig. 2(a) forthe numbering of the different locations.

124701-3 Torun et al. J. Appl. Phys. 116, 124701 (2014)

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:193.140.194.112 On: Mon, 29 Sep 2014 20:53:26

controlling phase noise of the oscillator since phase noise ofan oscillator is reduced with increasing quality factor.22 It ispossible to increase the quality factor further by decreasingthe size of the structure to reduce losses.

IV. BIOLOGICAL EXPERIMENTS

The response of the device for sensing applications canbe defined as the shift in resonant frequency of the SRRstructure with respect to changes in the sensing parameter.The most sensitive locations of the SRR structures to mea-sure dielectric loading are the edges of the slit (3rd and 4thlocations, see Fig. 2(a)). These correspond to the largest fre-quency shifts. We performed biomolecular experiments todemonstrate the use of the SRR-based device as a biosensor.We probed the interactions between fibroblast growth factor2 (FGF-2) and heparin. Custom prepared Murine recombi-nant FGF-2 and low molecular weight heparin molecules(Enoxaparin, Sanofi, Paris, France) are used in the experi-ments. FGF-2 plays important roles in biological processesincluding embryogenesis, angiogenesis, and wound healing.Heparin binds FGF-2 through a specific domain23 with highaffinity (dissociation constant of 22–30 nM).24

We started the experiment after s11 and s21 spectra ofthe device were captured. Then, we incubated the surfaceof parylene with FGF-2 molecules with a concentration of140 lg/ml. We placed a 10–ll droplet at location number 4and waited for 30 min at room temperature. Molecules wereanchored on the surface of the device. The incubated areawas uniformly coated due to relatively higher concentrationof FGF-2. Next, we dried the surface and placed a 20–lldroplet of heparin with a concentration of 10 lg/ml immedi-ately afterwards. The schematic of the experimental setup is

shown in Fig. 5(a). Then we dried the surface again andplaced heparin molecules with a concentration of 20 lg/mlby keeping the volume constant at 20 ll. We repeated thiscycle with increasing concentrations in heparin solution. Ateach step, we recorded s11 and s21 spectra as shown in Figs.5(b) and 5(c), respectively. For the final cycle, we placed adroplet of DI water for a control experiment.

Incubation of FGF-2 molecules with a 10–ll dropletincreases the effective permittivity of the device. Weobserved a 3.5% reduction in resonant frequency in thiscase. Adding heparin molecules reduces further the resonantfrequency. Heparin molecules bind to FGF-2 molecules onthe surface, increasing the permittivity. We observed a totalshift of 10% in resonant frequency after we added the hepa-rin solution with a concentration of 10 lg/ml. Resonant fre-quency of the device reduces with increasing molecularconcentration of heparin. On the other hand, the shift in reso-nant frequency of the device when we placed a droplet of DIwater by keeping the volume the same is almost identicalwith the case of FGF-2 incubation. This was a controlexperiment showing the specificity of the device response tomolecular interactions. Molecular interactions result ingreater shift in resonant frequency when compared to theloading effect of DI water that has large dielectric constant.

The parylene layer isolates the metallic rings from theelectrically conductive liquid containing biomolecules. Thepresence of parylene layer increases the effective permittiv-ity of the device and reduces its resonant frequency.However, the shift in the resonant frequency is determinedby the molecular concentration and is not directly affectedby the parylene layer.

Resonant frequency of the device with respect tochanges in molecular concentration is shown in Fig. 6. There

FIG. 4. Simulated profiles of (a) sur-face current density and (b) electricfield vectors of the SRR structure with-out any dielectric loading. The profilesof the SRR structure with a droplet ofDI water at 3rd location are shown in(c) and (d). See Fig. 2(a) for the num-bering of the different locations.

124701-4 Torun et al. J. Appl. Phys. 116, 124701 (2014)

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:193.140.194.112 On: Mon, 29 Sep 2014 20:53:26

is a linear dependency within the measurement range. Thelinear fit to the experimental results indicates a sensitivity of3.7 MHz/(lg/ml). The current experimental setup relies on anetwork analyzer for the measurement of shift in resonant

frequency of the device. The experimentally determined sen-sitivity value results in a resolution of sub-ng/ml levels witha frequency resolution of 1 kHz at the measurement fre-quency band.

V. CONCLUSION

In summary, we introduce an antenna-coupled split-ringresonator-based sensor for biological sensing applications.The device includes integrated antennas and operates atmicrowave frequencies. There is no need for external anten-nas to excite and measure the characteristics of the resonatorthanks to the integrated antennas. A thin film of parylene iscoated on the surface of the device to electrically seal the de-vice for in-liquid operation. Devices can be fabricated fastand at low cost using standard techniques for printed circuitboards with an additional step of parylene coating. We char-acterized the electromagnetic behavior of the device bymeasuring scattering parameters using a vector network ana-lyzer. The resonant frequency of the device is at 2.12 GHz.The characteristics of the device are suitable to develop anoscillator circuit to detect its resonant frequency. We charac-terized the effect of dielectric loading by placing droplets ofDI water at different locations over the surface of the device.Thus, we identified locations that result in largest shift in res-onant frequency. The results of experimental characteriza-tion are in good agreement with simulations. Finally, weperformed biomolecular experiments using the characterizeddevice by probing interactions between FGF-2 and heparin.We obtained the sensitivity of the device as 3.7 MHz/(lg/ml)with respect to changes in concentration of heparin.

ACKNOWLEDGMENTS

This work was supported by the T€ubitak Grant No.112E250. Dr. Tessa L€uhmann (University of W€urzburg,Germany) is acknowledged for supplying samples of Murinerecombinant FGF-2.

1V. V. Varadan and A. R. Tellakula, J. Appl. Phys. 100(3), 034910 (2006).2R. Liu, A. Degiron, J. J. Mock, and D. R. Smith, Appl. Phys. Lett. 90(26),263504 (2007).

3D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Schultz,Phys. Rev. Lett. 84(18), 4184 (2000).

4R. A. Shelby, D. R. Smith, and S. Schultz, Science 292(5514), 77 (2001).5D. Schurig, J. J. Mock, B. J. Justice, S. A. Cummer, J. B. Pendry, A. F.Starr, and D. R. Smith, Science 314(5801), 977 (2006).

6P. Gay-Balmaz and O. J. F. Martin, J. Appl. Phys. 92(5), 2929 (2002).7H.-T. Chen, W. J. Padilla, J. M. O. Zide, A. C. Gossard, A. J. Taylor, andR. D. Averitt, Nature 444(7119), 597 (2006).

8W. J. Padilla, M. T. Aronsson, C. Highstrete, M. Lee, A. J. Taylor, and R.D. Averitt, Phys. Rev. B 75(4), 041102 (2007).

9C. Rockstuhl, T. Zentgraf, H. Guo, N. Liu, C. Etrich, I. Loa, K. Syassen, J.Kuhl, F. Lederer, and H. Giessen, Appl. Phys. B 84(1–2), 219 (2006).

10A. N. Grigorenko, A. K. Geim, H. F. Gleeson, Y. Zhang, A. A. Firsov, I.Y. Khrushchev, and J. Petrovic, Nature 438(7066), 335 (2005).

11A. Ishikawa, T. Tanaka, and S. Kawata, Phys. Rev. Lett. 95(23), 237401(2005).

12H. Torun, S. Sadeghzadeh, and A. D. Yalcinkaya, Rev. Sci. Instrum.84(10), 106107 (2013).

13R. Melik, E. Unal, N. K. Perkgoz, C. Puttlitz, and H. V. Demir, Appl.Phys. Lett. 95(1), 011106 (2009).

14A. K. Horestani, C. Fumeaux, S. F. Al-Sarawi, and D. Abbott, IEEE Sens. J.13(4), 1153 (2013).

FIG. 5. a) Schematic of the experimental setup presenting the functionaliza-tion of the device with biomolecules. Typical (b) s11 and s21 spectra of thedevice measured during a biological experiment.

FIG. 6. The shift in resonant frequency of the device with respect to changesin the concentration of heparin.

124701-5 Torun et al. J. Appl. Phys. 116, 124701 (2014)

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:193.140.194.112 On: Mon, 29 Sep 2014 20:53:26

15W. Withayachumnankul, K. Jaruwongrungsee, A. Tuantranont, C.Fumeaux, and D. Abbott, Sens. Actuators, A 189(0), 233 (2013).

16M. S. Boybay, A. Jiao, T. Glawdel, and C. L. Ren, Lab Chip 13(19), 3840(2013).

17H.-J. Lee, J.-H. Lee, H.-S. Moon, I.-S. Jang, J.-S. Choi, J.-G. Yook, andH.-I. Jung, Sens. Actuators B 169(0), 26 (2012).

18H.-J. Lee and J.-G. Yook, Appl. Phys. Lett. 92(25), 254103 (2008).19H.-J. Lee, H.-S. Lee, K.-H. Yoo, and J.-G. Yook, J. Appl. Phys. 108(1),

014908 (2010).

20H. Torun, K. K. Sarangapani, and F. L. Degertekin, J. Microelectromech.Syst. 19(5), 1021 (2010).

21O. Sydoruk, E. Tatartschuk, E. Shamonina, and L. Solymar, J. Appl. Phys.105(1), 014903 (2009).

22D. B. Leeson, Proc. IEEE 54(2), 329 (1966).23M. Kan, F. Wang, J. Xu, J. W. Crabb, J. Hou, and W. L. McKeehan,

Science 259(5103), 1918 (1993).24H. Rahmoune, H.-L. Chen, J. T. Gallagher, P. S. Rudland, and D. G.

Fernig, J. Biol. Chem. 273(13), 7303 (1998).

124701-6 Torun et al. J. Appl. Phys. 116, 124701 (2014)

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:193.140.194.112 On: Mon, 29 Sep 2014 20:53:26