size‐selective analyte detection in an biosensor...size‐selective analyte detection in an...

Size‐selectiveanalytedetectioninanintegratedopticalYounginterferometer

biosensor

HarmenK.P.Mulder

Membersofthedissertationcommittee:Prof.dr.ir.J.W.M.Hilgenkamp UniversityofTwente(chairmanandsecretary)Prof.dr.V.Subramaniam UniversityofTwente(promotor)Dr.ir.J.S.Kanger UniversityofTwente(assistantpromotor)Prof.dr.K.J.Boller UniversityofTwenteProf.dr.J.C.T.Eijkel UniversityofTwenteDr.ir.C.Blum UniversityofTwenteProf.dr.H.Salemink RadboudUniversityNijmegenProf.dr.L.M.LechugaGómez InstitutCatalàdeNanociènciaiNanotecnologia

Cover:Schematicrepresentationofacrosssectionofawaveguidewithpropagatingmodes of three different wavelengths which can be used to discriminate betweendifferentsizedsubstances.Theworkdescribed in this thesiswas carriedout atNanobiophysics group,MESA+Institute for Nanotechnology, Faculty of Science and Technology, University ofTwente, P.O. Box 217, 7500 AE Enschede, The Netherlands. This thesis is part ofNanoNextNL,amicroandnanotechnologyinnovationconsortiumoftheGovernmentoftheNetherlandsand130partners fromacademiaandindustry.More informationonwww.nanonextnl.nl.

HarmenKlaasPeterMulderSize‐selective analyte detection in an integrated optical Young interferometerbiosensorPh.D.thesis,UniversityofTwente,Enschede,TheNetherlandsPrintedby:GildeprintDrukkerijen–EnschedeISBN:978‐90‐365‐4029‐2DOI:10.3990/1.9789036540292Copyright©2016byH.K.P.MulderAllrightsreserved.

SIZE‐SELECTIVEANALYTEDETECTIONINANINTEGRATED

OPTICALYOUNGINTERFEROMETERBIOSENSOR

PROEFSCHRIFT

TerverkrijgingvandegraadvandoctoraandeUniversiteitTwente,

opgezagvanderectormagnificus,Prof.dr.H.Brinksma,

volgensbesluitvanhetCollegevoorPromotiesinhetopenbaarteverdedigenop

vrijdag19februari2016om12.45uur

door

HarmenKlaasPeterMuldergeborenop30‐10‐1985

teSneek

Ditproefschriftisgoedgekeurddoor:Prof.dr.VSubramaniampromotorDr.ir.J.S.Kangerassistantpromotor

v

TableofcontentsChapter1Introduction............................................................................................................1

1.1Whatisabiosensor?.........................................................................................................................2

1.2Biosensorcriteria..............................................................................................................................2

1.3Label‐freeintegratedopticalbiosensors.................................................................................5

1.4Methodsforimprovingspecificity..........................................................................................17

1.5Outlineofthethesis.......................................................................................................................18

Acknowledgements...............................................................................................................................18

References.................................................................................................................................................19

Chapter2Size‐selectivedetectioninintegratedopticalinterferometricbiosensors..................................................................................................................................23

2.1Introduction......................................................................................................................................24

2.2Theoreticalaspects........................................................................................................................25

2.3Resultsanddiscussion..................................................................................................................30

2.4.Conclusions.......................................................................................................................................37


Appendix2.AChromaticdispersion..............................................................................................38

Appendix2.BDerivationsensitivitycoefficient........................................................................39

Appendix2.CDerivationrelativeprecision................................................................................40

Appendix2.DDerivationsurfacemasscoverage.....................................................................41

References.................................................................................................................................................43

Chapter3Design,realizationandcharacterizationofasize‐selectiveYounginterferometersensor............................................................................................................45

3.1Introduction......................................................................................................................................46

3.2Sensingplatform.............................................................................................................................47

3.3Lightsources.....................................................................................................................................49

3.4Incoupling..........................................................................................................................................50

vi

3.5Imaging................................................................................................................................................56

3.6Detection.............................................................................................................................................58

3.7Dataprocessing...............................................................................................................................65

3.8Overviewsetup................................................................................................................................70

3.9Characterizationofphasenoiseanddrift............................................................................71


Appendix3.AFocalpositionusinga4flenssystem...............................................................74

References.................................................................................................................................................75

Chapter4Differentanalysisapproachesforsize‐selectiveanalytedetection....77

4.1Introduction......................................................................................................................................78

4.2Theoreticalanalysisapproach..................................................................................................78

4.3Ratio‐basedanalysisapproach.................................................................................................79

4.4Combinedanalysisapproach.....................................................................................................82

4.5InfluenceinputparametersinanalysisapproachesondeterminedRIchange..85

4.6Influenceofnoise,driftandartefactsondeterminedRIchange...............................89

4.7Discussionandconclusions........................................................................................................92


Appendix4.ASlightlydifferentmatrixforratio‐basedapproach....................................93

References.................................................................................................................................................94

Chapter5Experimentalapplicationofanalysisapproachesforsize‐selectiveanalytedetection.....................................................................................................................95

5.1Introduction......................................................................................................................................96

5.2Reliabilityofsensor.......................................................................................................................97

5.3Discriminationbetweentwodifferentsubstances..........................................................98

5.4Discriminationbetweenthreedifferentsubstances.....................................................102

5.5Simultaneoususeofothertechniquesnexttosize‐selectivedetection...............107

5.6Blindexperimentwith85nmbeadsandproteinA......................................................109

5.7Validationandreproducibilityofsize‐selectivedetection.........................................111

vii

5.8Conclusions......................................................................................................................................114

Acknowledgements.............................................................................................................................115

Appendix5.ACleaningprotocol....................................................................................................116

Appendix5.BRelationbetweenappliedbeadconcentrationandmeasuredsurfacemasscoverage........................................................................................................................................116

References...............................................................................................................................................120

Chapter6Size‐selectiveanalytedetectionusingmultiplewavelengthsandpolarizations..........................................................................................................................121

6.1Introduction....................................................................................................................................122

6.2Experimentalrealisation...........................................................................................................123

6.3Proof‐of‐principleexperiments..............................................................................................124

6.4Discussionandconclusions......................................................................................................130


References...............................................................................................................................................132

Chapter7Timingdifferenceofrefractiveindexchangesinducedbybindingofproteinsandbulkchanges.................................................................................................133

7.1Observedtimedelayofbulksignalcomparedtobinding..........................................134

7.2SimulationsD‐glucoseandproteinAinsensingwindow...........................................137

7.3TimedelayBSAandD‐glucose................................................................................................144

7.4Timedelayasfunctionoftubelength..................................................................................146

7.5Conclusionsandforwardlook................................................................................................148


References...............................................................................................................................................149

Chapter8Applicationsforsize‐selectivedetection:aforwardlook...................151

8.1Introduction....................................................................................................................................152

8.2Technologyassessment.............................................................................................................153

8.3Conclusions,reflectionsandforwardlook........................................................................164


viii

Appendix8.AWorkshoppreparations.......................................................................................166

Appendix8.BPicturesworkshop..................................................................................................169

References...............................................................................................................................................170

Listofabbreviations............................................................................................................171

Summary.................................................................................................................................173

Samenvatting.........................................................................................................................179

Listofpublications...............................................................................................................185

Dankwoord.............................................................................................................................187

Chapter1

Introduction

AbstractThis thesis presents a Young interferometer (YI) biosensor that can perform size‐selective measurements using multiple wavelength excitation. This size‐selectivedetection can be used to improve the specificity of evanescent field‐based opticalbiosensorsandisbasedonthevarioussensitivitiesoftheevanescentfieldsofmultiplewavelengths. The approach of using multiple wavelengths is, in addition to YIbiosensors,alsoapplicabletoothertypesofevanescentfield‐basedopticalsensors.Inthis chapter, we first introduce the general concept of biosensors, followed by adiscussion of important criteria for biosensors. The most widely‐used evanescentfield‐basedoptical sensorsarereviewed. The limitingspecificityof thesesensors isaddressed, togetherwith techniques to improve the specificity. Finally,we proposethe new approach of size‐selective detection, based on the use of multiplewavelengthstoimprovethespecificityofevanescentfield‐basedopticalbiosensors.

2 Chapter1

1.1Whatisabiosensor?

Asensorcanbedefinedasadevicethatdetectsachange inaphysicalstimulusandconverts this into a measurable signal. If a sensor is used for detection andquantificationofthedetectedmaterialinabiologicalsamplewespeakofabiosensor.The detectedmaterial, which is called the analyte, could, for example, be a diseasebiomarker,anenzyme,avirusoraprotein.Biosensorsgenerallyconsistofasensitiverecognition element that recognizes a specific analyte and as a result produces aphysicochemicalsignal.Enzymes,antibodies,nucleicacids,cellreceptors,tissue,andmicroorganisms are several examples of biological elements used in biosensors.Furthermore,thebiosensorcontainsatransducerelementthatconvertsthedetectedphysicochemicalsignalintoameasurablesignal.Theintensityofthissignalisdirectlyor inversely proportional to the analyte concentration in themeasurement sample.Fig.1.1showsaschematicoverviewofabiosensor.

Fig.1.1:Schematicoverviewofabiosensor

1.2Biosensorcriteria

Dependingonitspurpose,abiosensormustmeetseveralcriteria.If,forexample,thepurposeofthebiosensoristomeasureverylowconcentrationsofaspecificanalyte,thebiosensormustbeverysensitiveandspecific.Othercriteriaareimportantifhighconcentrationsoftheanalyteneedtobemeasuredrapidlyandinexpensively.Thesixmaincriteriaforbiosensorsare:

Sensitivity Specificity Scalabilitytosmallerdimensions Measurementtime Measurementcosts Multiplexing

andarediscussedhere.

Introduction 3

1.2.1SensitivityThemoresensitivethebiosensor,thelowerthemeasurableconcentrationofanalyte.Thesensitivityofthebiosensorisstronglydeterminedbythetransducerelementofthe biosensor. Widely used transducers in biosensors are optical, electrochemical,micromechanical, thermal or piezoelectric. However, optical transducers play apredominant role, because of their high sensitivity and high bandwidth [1].Additionally,opticaltransducershavetheadvantagesofbeingnon‐invasiveandnon‐destructive, free of electrical or explosion risks and immune to electromagneticinterference.Therefore,wefocusonopticaltransducers.

Apart fromthe typeof transducerelement, the sensitivityofabiosensor is alsodetermined by the recognition element of the sensor. Depending on which type ofelement is used, biosensors can be classified as labelled or label‐free. In a labelleddetection scheme, a label (for example fluorophores, enzymes or radionuclides) isusually attached to the analyte to allow sensing. The main advantage of labelledbiosensors is their high potential to detect low concentrations. With fluorescence‐baseddetection, the sensitivity canevengodown to singlemolecules [2].However,labelshavetheirown inherentproblems; forexample, fluorophorescanquenchandphotobleach, and radioisotopes have a short lifetime, high costs and can producehazardous contaminants. Furthermore, quantitative measurements are challengingwhen using labelled sensing, because it is difficult to control the exact number oflabelsoneachmolecule [3].Most importantly, the labelhas tobe conjugated to therecognitionelement,whichrequiresfurthersamplehandingsteps,sonodirecton‐sitemeasurement ispossible.Conversely, label‐freeopticaldetection ishighlysuited forkinetic and quantitative measurements of molecular interactions. For label‐freeopticaldetection,abioreceptorlayerisusuallyused.Thesensitivityoftheselabel‐freeoptical sensors is also determined by the specificity of this bioreceptor layer, thestability of the linker between the molecule and the surface, and the number ofavailable binding sites. In conclusion, the selection of the recognition element iscrucialfordeterminingthesensitivityofthebiosensor.

Thesensitivityofabiosensorcanbeexpressedasalimitofdetection(LOD).TheLOD of an optical biosensor can be expressed in refractive index units (RIU) or insurfacemassdensity(pg/mm2)foranybiosensorthatissensitivetoanyaccumulationofmassonitssurface.

1.2.2SpecificityThe specificity of a biosensor is ameasure for how specific themeasured signal iswhendetectingtheanalyte.Thespecificityofabiosensorismainlydeterminedbythebiologicalelementofthebiosensor.Forlabelledbiosensorsitisimportanttocorrectlylabeltheanalyte,soasuitablelabelshouldbefoundforeachanalyte.Thelabelmustnot have any effect on the function of themolecule andmust not block the active

4 Chapter1

binding sites of the molecules. On the other hand, for label‐free detection, thebioreceptor layermust be highly specific toward the analyte andmust have a highaffinity towardtheanalyte.Furthermore, the interactioneventbetweentheelementandanalytemustbedetectablebythetransducerelement.Proteins,nucleicacidsandantibodiesaremostcommonlyusedasbioreceptorlayers.Suchbioreceptorlayerscanbe immobilized at the surfaceon thebasis ofphysical adsorption, covalentbinding,non‐covalent interactions to a previously deposited layer (for example, biotin‐streptavidin or protein A for antibodies), physical entrapment, and self‐assembledmonolayers.Itisimportantthatthespecificityandaffinityoftheelementmustnotbealtered significantly by its immobilization on the surface of the optical transducer.Referencechannels,blockingtechniquesandwashingstepsareusedto improvethespecificity of label‐free biosensors, as the specificity of the bioreceptor layer is notalways sufficient to accurately determine the analyte concentration. Referencechannelsareusedtocanceloutbulkeffects,whereblockingtechniquespreventnon‐specificbindingandwashingstepsareusedtoremovenon‐specificboundparticles.

1.2.3ScalabilitytosmallerdimensionsLab‐on‐a‐chip(LOC)devicesareminiaturizeddevicesinwhichall functionalitiesareintegrated on the same platform and that offer significant advantages overconventional analytical methods [4]. Such LOC devices have to be both small andportable, and therefore the biosensor should preferably be scalable to smalldimensions.Forthatreason,integratedoptical(IO)devicesofferanidealsolutionforLOC platforms [1]. Apart from the advantages for all optical sensors, IO devicescombinemechanicalstabilitywithscopeforminiaturization.Besidestheadvantagesof being small and portable, smaller devices also lead to reductions in volumes ofsampleandreagentsrequired.

1.2.4MeasurementtimeThemeasurementtimeisalsoimportantforanybiosensorthathastodetectanalytesrapidly and in real‐time. Many techniques, such as enzyme‐linked immunosorbentassay(ELISA)[5,6]andpolymerasechainreaction(PCR)[7],areverysensitiveandspecific, but require multiple processing steps and therefore considerable time toperform.Label‐freebiosensorsusuallyrequiremuch lesstimetoperformcomparedto labelled biosensors, because no time is needed for labelling the analyte,amplificationstepsorpurificationofsamples.Moreover,label‐freetechniquescanbeused tomeasure in real‐time andon‐site, because analytes canbedetected in theirnatural forms. If the label‐free techniques are combined with IO devices thatincorporatemicrofluidics, the response time of the biosensor can be reduced evenfurther.

Introduction 5

1.2.5MeasurementcostsBiosensorsshouldpreferablybeasinexpensiveaspossiblewithoutcompromisingtheessential properties of the biosensor. Required labels and sometimes laboriouslabellingprocessesrepresentasubstantialpartofthecostsoflabelledbiosensors.Onthe other hand, label‐free techniques entail the costs of a bioreceptor layer. TheadvantageofIOopticaldevicesisthattheycanbebothsmallandmass‐produced,andthereforehavethepotentialtobeinexpensive.

1.2.6MultiplexingMultiplexingistheabilitytosimultaneouslymeasurevariousanalytesfromthesamesample. Optical transducers have a potential for parallel sensingwhichmeans thatthey can be used for multiplexing [1, 8]. Nevertheless, multiplexing for labelledbiosensors canbe complicated, as it requiresmultiple labels and suitabledetection.On the other hand, label‐free detection needs suitable bioreceptor layers. However,label‐free detection can be combined with IO devices, which are highly suited formultiplexing,becauseof thepossible fabricationof arraysof sensorswith the samecharacteristics within the same chip and their great flexibility in the choice ofmaterialsandthestructuresselection[4].

1.3Label‐freeintegratedopticalbiosensors

Basedonthepreviouslydiscussedselectioncriteria,thefocusinthisthesiswillbeonlabel‐free IO detection. Most IO sensors rely on the evanescent field detectionprinciple. In such sensors, the electromagnetic waves, called guided modes, areconfined in awaveguidedue to total internal reflection (TIR).However, part of theelectromagneticwaves,theevanescentfield,penetratestheinterfacesofthedifferentmaterials of the waveguide. Interaction between the analyte and the bioreceptorcoatedon thewaveguidesurfaceresults inachange inrefractive index(RI).This inturnchangesthepropagationoftheguidedmodethroughthewaveguidebychangingthe evanescent field. The evanescent field‐based sensors can be classified as non‐linearwhenguidedmodesaregeneratedthathaveadifferentwavelengthto thatofincident light(e.g.Ramandetection).Linearevanescentfield‐basedsensorsmeasurethechangeintheRIintheevanescentfieldasamodificationoftheintensity,phaseorpolarizationof theoutputsignal.The linearevanescent field sensorscanbedividedinto two categories, bothmeasuring the RI change in the evanescent field. On onehand,therearetheabsorptionsensors, inwhichthechangeintheimaginarypartoftheRIresults inamodificationofthe light intensityat theendofthedevice.Ontheotherhand, thechangescanalsobecreated in the realpartof theRI, resulting inachange in the propagating velocity of the guided mode. We focus on the linear

6 Chapter1

evanescentfield‐basedsensor,basedonthedetectionofchangesintherealpartoftheRI. The most important examples of this type of sensor are reviewed in this sub‐section.AsacomparisontotheseIOsensors,thecommonly‐usedopticalevanescent‐fieldbasedbiosensor,SurfacePlasmonResonance(SPR),isadded.Table1.1onpage17showsacomparisonoftheLODofvariousbiosensors.

1.3.1SurfacePlasmonResonancesensorsTheSPRsensor isawidelyusedtypeofbiosensor.Surfaceplasmonscanbeexcitedwhen an incident beam of transverse magnetic (TM)‐polarized light strikes a thinelectricallyconductinglayeratinterfaceofathinmetalfilmandadielectricmaterial.UnderconditionsofTIR,theincidentlightwillbecoupledtoasurfaceplasmon(SP)iftheincidentlightwavevectorcomponentparalleltothesurfacematchesthesurfaceplasmonwavevector(ki=kp).Theincidentlightwavevectorcomponentparalleltothesurface,ki,isgivenby:

2

sini p ik n

, (1.1)

wherethewavelengthoftheincidentlight,nptherefractiveindexoftheprismandi theangleoftheincidentlight.TheSPwavevector,kpisexpressedas:

1 2

1 2

1 2

2pk

, (1.2)

where, ε1 and ε2 are the dielectric permittivity constants of themetal film and thedielectric exit medium, respectively. Coupling is performed by a coupling device.Grating couplers,waveguide couplers and prims couplers are examples of couplingdevices.OnecanspeakoftheKretschmannconfigurationifthereisathinmetalfilmatthesurfaceoftheprism,asshowninFig.1.2a.Abioreaction,causinganRIchangeinthedielectricand therebya change inε2, results ina change to theSPwavevector.Thiscanbequantifiedbymeasuringthechangesinthecharacteristicsofthereflectedlight, commonly achieved bymeasuring the coupling angle, couplingwavelength orintensity. Depending onwhich variable ismeasured, the SPR sensor is classified intermsofangular,wavelength, intensityorphasemodulation [9].ThoseSPRsensorsbased on wavelength, intensity or phase modulation (see Fig. 1.2b) provide thehighest RI resolutions [10]. The best SPR sensor based on prism coupling with alimited number of channels (<10) provides a RI resolution of 10‐7 RIU [10]. SPRsbasedongratingcoupling typicallyachieveresolutionsof10‐5‐10‐6RIU [11,12]andwaveguide based SPRs, typically achieve resolutions of 10‐5‐10‐6 RIU [13, 14]. Aresolution of 3 x 10‐8 RIU has also been reported [15]. However, these long‐rangesurface plasmons, which are called long‐range because of their low attenuation, so

Introduction 7

propagation over centimetres, penetrate much further (1400 nm vs. 200 nm forconventionalSPR)intothesensingregion;thereforethesensorcanonlybefullyusedwhenlargeanalytesaretargetedorbiorecognitionelementsare immobilizedonthesurfaceandcombinedwithanextendedcouplingmatrix[10].Averagemassdetectionlimits of 1‐5 pg/mm2 were reported [10]. The SPR biosensor is successfullycommercialised by GE Healthcare as Biacore™ [16]. SPR is widely used due to itsrobustness and simplicity. However, its large size makes it complex to achieveminiaturisationsuitableforLOCdevices.

Fig.1.2: a) SPR in Kretschmann configuration based on b) angular, wavelength or intensitymodulation.

1.3.2GratingcouplersA grating coupler consists of a single‐mode waveguide containing periodicdisturbances. At a certain angle, at which the incoupling condition is fulfilled, thegratingpermitstheexcitationofaguidedmode.Theincouplingconditionisgivenby:

sineff air iN n q , (1.3)

whereNeff is theeffective refractive indexof thewaveguide,nair theRIofair, i the

angle of the incident light,q the diffraction order, thewavelength of the incidentlightandΛthegratingperiod.IftheRIchanges,forexampleduetoabioreaction,thisresults in a change in Neff. As a consequence, the optimal coupling angle changes.Photodetectorsattheendofthewaveguidemeasuretheintensityofthecoupledlightas a function of the angle i , which angle is changed using a rotation stage with

precisemechanicalmovement. Alternatively, the outcoupling angle can be used forsensing, which does not require a rotation stage [17]. Both the incoupling andoutcoupling configuration show an LOD of 10‐6 RIU. Furthermore, a setupconfiguration based on reflection‐mode operation has been developed [18], which

8 Chapter1

avoidsmovingpartsinthesetup.Alensisusedtosimultaneouslyirradiatearangeofangles onto the grating. Due to the effective coupling into guidedmodes under thecouplingangle, aminimumof the reflected light isobservedand followedbyaCCDcamera.Applyingthismethod,anLODof3x10‐6RIUisreported.Besides,aso‐calledwavelength interrogated optical sensor (WIOS) based on grating couplers has beendeveloped [19, 20]. One grating coupler is used for coupling the light into a single‐modewaveguide.Theothergrating isused for lightoutcoupling.Ata fixedangleofincidence,whichcanbesetusinganadjustablemirror,theRIchangesaremonitoredby scanning the resonance peak, by using a tuneable laser diode. At the resonancewavelength, light is coupled into the waveguide. Next, the light is coupled out viaanother grating and collected by a multimode fiber, which directs the light to aphotodiode.AnLODof <10‐6RIUwasreportedandbymeasuringsmallmolecules,amasscoverageof100fg/mm2wasobtained,correspondingtoadetectionlimitof0.3pg/mm2[19].Gratingcouplersaresuitableformultiplexingandduetoitssimplicityand ability to measure label‐free, the technology has been commercialised byMicroVacuumLtd [21]. Fig.1.3 showsa schematicoverviewof two typesofgratingcouplersensors.

Fig. 1.3: Two types of grating couplers, both using a grating to couple the light into thewaveguide,whereoneusesadetectorattheendofthewaveguidemeasuringtheintensityasafunctionofthecouplingangleandtheotherusesasecondgratingtocouplethelightoutofthewaveguide,measuringthechangeinresonantwavelength.

1.3.3RingresonatorsensorsInringresonatorsensors,lightiscoupledintoacircularwaveguideviatheevanescentfieldofaninputwaveguide.AsaresultofTIRalongthecurvedboundary,thewavespropagate through thewaveguide in the formofwhispering gallerymodes (WGM).Thesensitivityisincreasedbyconstructiveinterferenceoverthemultipleround‐trips.TheWGMcirculatesalongtheresonatorsurfacemanytimeswhere it interactswiththe analyte via the evanescent field. In contrast with straight waveguides, theinteractionisnolongerdeterminedbythelengthofthewaveguide,butbythenumber

Introduction 9

of light circulationswithin the ring,which is characterizedby the resonator qualityfactor(Q‐factor).TheeffectivelengthLeff,isgivenby:

2eff

QL

n

, (1.4)

whereistheresonantwavelengthandnistheRIoftheringresonator.Q‐factorsof106 can be reached in resonators resulting in an Leff of several centimetres.. ThedetectionisbasedonRIchanges,whicharerelatedtotheresonantwavelengthby:

2 eff

rN

M, (1.5)

whereNeffistheeffectiverefractiveindexexperiencedbytheWGM,rtheradiusofthering, and M is an integer describing the WGM angular momentum. Therefore, abioreaction–causinganRI changenear thesurfaceof the ring resonator–changesthe effective RI, leading to a spectral shift in theWGM. This can be monitored byscanningthewavelengthorbymeasuringtheintensityprofileatafixedwavelength.Fig.1.4showsaschematicoverviewofaringresonatorsensor.Ringresonatorshavethepotentialtobecombinedintohighlydensearrays,whichisavaluablefeatureformultiplexing. Sensitivities of 7.6 x 10‐7 RIU [22] and 1.5 pg/mm2 [23] have beenachieved and the Maverick Detection System ring resonator sensor has beencommercializedbyGenalyte[24].

Fig.1.4:Schematicoverviewofaringresonatorsensor,inwhichRIchangesaremeasuredasachange in transmission wavelength, as the light coupled from the input waveguide to thecircularwaveguidechanges,becauseof thechange in theresonantwavelengthof thecircularwaveguide.

10 Chapter1

1.3.4PhotoniccrystalbasedsensorsPhotoniccrystal(PC)basedsensorsarewell‐definednanostructures,whichconsistofaperiodicityintheorderofthewavelengthofthelight.Photonicbandgaps(PBG)aregenerated;thelightwhosewavelengthlieswithinthePBGcannotpropagatethroughthe PC. This results in awide stopband in the transmission or reflection spectrum.However,bylocallydisturbingthestructureofthePC,aphotonic“defect”withinthebandgapcanbeintroduced,leadingtotheformationofadefectmode.Consequently,lightresonantwiththisdefectmodecanpropagatethroughthePC,resultinginapeakwithinthebandgap.Thespectralpositionofthispeakstronglydependsonthe localenvironment around the defect. For that reason, this can be used for sensingmoleculesbindingtothedefect,whichisillustratedschematicallyinFig.1.5.Paralleldetectionandlightinteractionwithvolumesdowntofemtolitresarepossible,becauselight canbe localizedandconcentrated in very small volumes. LODsof7 x10‐5RIU[25],2.4x10‐3RIU,4.9x102pg/mm2[26]and2.1pg/mm2[27]havebeenachieved.TheseLODswereachievedwithtypicallyachievedsensitivitiesofaround100–175nm/RIU,whereLiuetal.showasignificantlyhighersensitivityof460nm/RIUandadesignwithmultiplepeaks in the transmission spectrawhichenables simultaneousdetection of analytes [28]. Kita et al. presented an LOD of 9.0 x 10‐5 RIU with anexpectationofaresolutionof<10‐6RIU[29].AsthePCbasedsensorsarequiterecent,compatible microfluidics, surface chemistry and detection procedures to minimizenon‐specific binding still need to be optimized. Nevertheless, it was possible tospecificallydetectproteinwithPC’s[30].

Fig.1.5:Schematicoverviewofaphotoniccrystalbasedsensor,wherebindingofananalytetotheantibodyresultsinashiftintransmissionwavelength.

1.3.5InterferometricwaveguidesensorsInterferometricwaveguide biosensors are themost attractive of various IO sensorsdue to theirhighsensitivityandbroaddynamicrange [4].Next to theiroutstandingsensitivity, advantages of the interferometricwaveguides are the cost‐effective andsimpleproductionand the abilityofmultiplexing andminiaturization [31]. Twoormoreconfinedlightwavesformaninterferencepatternwhichismeasuredovertime.

Introduction 11

Interactionswiththeexternalmedium(sample)viatheevanescentfield(see.Fig.1.6)result in changes in velocity between the waves, which can be analysed from theinterferencepattern.Highsensitivitiescanbeachieved,duetotheoptionofusinglonginteraction lengths. Mach‐Zehnder, bimodal, dual polarization, Hartman and Younginterferometersareexamplesof interferometricdevicesusedasbiosensorsandaredescribedbelow.

Fig.1.6: Schematic overview of a typical side view of a waveguide used for interferometricwaveguide sensors, where the binding of the analyte to the antibody results in a change invelocityoftheguidedmode(redline),whichcanbeanalysedfromaninterferencesignal.Theevanescentfield,whichisguidedmodethatpenetratesintothesample,isusedforsensing.

Mach‐ZehnderinterferometerInaMach‐Zehnderinterferometer(MZI),coherentlightiscoupledintoasingle‐modewaveguide. A single‐mode waveguide is required as higher order modes responddifferently than single‐modeonRI changes in the evanescent field. If a single‐modeand higher order modes simultaneously propagate through the waveguide, theirdifferentsensitivitiesresult inamisleadingsignal.AschematicoverviewofaMach‐Zehnder setup is shown in Fig. 1.7. Via a Y‐junction the light is split up into twochannels: one reference arm and onemeasuring arm. A bioreaction on the sensingareawithintheevanescentfieldofthemeasuringarmwillcauseachangeinNeff.Thisinducesaphasechangeandconsequentlyachangeintheinterferencewhenthetwoarms are recombined. At the output of the sensor, a photodiode measures theintensityIt,whichisexpressedas:

1 22 cos ( )t s r s rI I I I I t , (1.6)

whereIsistheintensityofthesignalarm,Iristheintensityofthereferencearmand thephasechangedescribedby:

12 Chapter1

, ,

2( ) eff s eff rt l N N

, (1.7)

wherethewavelengthoftheincidentlight,lthedetectionlengthandNeff,sandNeff,rtheeffective refractive indexof thesignalandreferencearms, respectively.Due thecosine dependency of the intensity, the sensitivity depends on the position of theinterferometriccurve.Thesignalchangeislowerneartheminimaandmaximaofthecosinecomparedtothequadraturepoints(pointswhere ( 1/2)m ,withm=

1,2,3…)ofthecosine.However,ifthequadraturepointsmovebothcontinuouslyandlinearlyduringa sensingevent, thequadraturepoints canbe tracked insteadof theoutputintensity.Followingthisimprovement,anLODof5x10‐8RIUwasreportedbyHeidemanandLambeck[32].OtherLODsreportedare1.9x10‐7RIUbyDanteetal.[33]and1x10‐7RIUbyZinovievetal. [34].MZI’swerealsocombinedwithgratingcouplersbyKozmaetal.[35]resultinginanLODofsurfacemassdensityof1pg/mm2andbyDuval etal. [36] resulting inanLODof1.6x10‐7RIU.Lambeck reportedanLODofsurfacemassdensityof0.01pg/mm2[37].TheMZIiscommercialisedintoabiosensor byOptisense [38] and Creoptix GmbH [39]which also combines theMZIwithgratingcouplers.

Fig.1.7:SchematicoverviewofaMach‐Zehnderinterferometer,inwhichananalytebindingtoanantibodycausesachangeinvelocityinthemeasuringarmcomparedtothereferencearm,resultinginachangeintheinterferencesignal.

Introduction 13

HartmaninterferometerTheHartmaninterferometer[40]usesagratingtocouplelinearlypolarizedlightintoa singlemodewaveguide.Next, the lightpropagates throughbothameasuringarmand a reference arm. Again, the evanescent field is used for sensing a bioreaction.Afterpassingthroughthesensingregioninthemeasuringarmthelightiscombinedwith the light from the reference arm by an array of optical elements, creating aninterferencepattern.Finally,theinterferencesignaliscoupledoutbyanothergratingintoaphotodiode.Fig.1.8showsaschematicoverviewoftheHartmaninterferometer.AHartmaninterferometerbasedsensorachievedanLODof10‐6RIU[41].Themainadvantagesof theHartmantechniqueare itssimplicityandtheability to implementmultiplexing without increasing the complexity of the setup. However, additionaloptical elements,whichare required formultiplexing, andoff‐chipdetection canbedrawbacksforLOCdevicesbasedonHartmaninterferometers.

Fig.1.8: Schematicoverviewof aHartman interferometer, inwhichananalytebinding toanantibodychangesthevelocityinthemeasuringarmcomparedtothereferencearm,causingachangeintheinterferencesignal.

BimodalwaveguideAbimodalwaveguide (BiMW) is also based on interferometry.However, instead ofusingtwoarms(asignalarmandareferencearm),asinglewaveguide isused.Thesinglewaveguideconsistsoftwodifferentzones:astartingsingle‐modesectionandasecondbimodalsectionthatacceptstwomodes;forexample,azero‐ordermodeandafirst‐ordermodewhichformaninterferencepattern.RIchangescanbesensedbytheevanescent field in the sensing region which is located within the cladding. A

14 Chapter1

bioreactioncausingsuchanRIchangecanbedetectedasachangeintheinterferencepattern.Theconfigurationofthistypeofinterferometer(seeFig.1.9)islesscomplexthanforexampletheMZIinterferometer.However,thereportedsensitivitieswiththebimodal interferometer of 2‐4 x 10‐7 RIU [42‐44], are lower compared to thoseachievedusingtheMZIinterferometer.

Fig.1.9: Schematic overview of a bimodal interferometer, inwhich an analyte binding to anantibodyresultsinvariouschangesofthevelocitiesofthedifferentpropagatingmodes,causingachangeintheinterferencesignal.

DualpolarizationinterferometerThedualpolarization interferometer(DPI)consistsofastructureof five layerswithvariousRI,oneon topof theother.Fig.1.10 isa schematicoverviewof theDPI.BymeansofTIR,thelightisconfinedtothesecondandfourthlayerswhichbothhaveahigher RI than the first, third and fifth layers. These twowaveguides are used as areferenceandasensingarm.TheevanescentfieldofthesensingarmpenetratesintoasensingregionandisusedforthedetectionofRIchanges,inducedbyabioreactionforexample. The light propagating through the reference arm remains constant, as theevanescent fielddoesnotreachthesurfaceof thedevice that isusedas thesensingregion.Bothwaveguidesareexcitedsimultaneouslyandat theoutputof thedevice,the lightwill forman interferencepatternwhich isdetectedbyaphotodiodearray.The polarization of the laser is modulated over time in order to excite TM andtransverse electric (TE) modes. The change in propagation velocity due to abioreaction isnot thesame forbothpolarizations.Therefore, theRI changeand thechangeinthicknessofaproteinlayercanbedeterminedwithoutambiguityandreal‐

Introduction 15

time [45].TheDPIwas commercialisedasAnaLight® interferometerby theFarfieldGroupwhocanreachanLODof10‐7RIU[4]andasurfacemassdensityLODof<0.1pg/mm2[46].

Fig.1.10:Schematicoverviewofadualpolarizationinterferometer,inwhichananalytebindingtoanantibodycausesachangeinvelocityinthemeasuringarmcomparedtothereferencearm,resultinginachangeinvariousshiftsoftheinterferencepatternsofthetwopolarizations,thatarethenusedtodeterminetheRIchangeandthicknesschangeofaproteinlayer.

YounginterferometerTheYI isanotherwidelyusedbiosensor. InaYI, light iscoupled intoasinglemodewaveguidestructure.Next,ay‐splitterisusedtosplitupthelightintotwowaveguidearms,whichareusedas reference armand sensing arms. Subsequently, the light iscoupledoutofthechip,afterwhichitwillrecombineandformaninterferencepatternwhich is then detected by a charge‐coupled device (CCD) camera. A schematicoverview is shown in Fig. 1.11. The evanescent field senses RI changes near thesurface of the sensing arm. Thismeans that a bioreaction that causes anRI changeresultsinaneffectiverefractiveindexchange effN ofthesensingarmwithrespectto

the reference arm. Consequently, the interference pattern changes, fromwhich thefollowingphasedifferencebetweenthedifferentinterferingbeamsisgivenby:

, ,

2eff s eff r

d yy N N l

L, (1.8)

16 Chapter1

where is the wavelength of the incident light, d the distance between the twochannels,Lthedistancebetweenoutputofthesensorandthecamera,ythepositionofthecamera,lthedetectionlengthandNeff,sandNeff,rtheeffectiverefractiveindexofthesignalarmandreferencearm,respectively.IftheRIchangesinthesignalarm,thisresultsinachangeinNeff,s,resultinginphasechangewhichismeasuredfromapplyingaFastFourierTransform(FFT)ontheinterferencepatternandreadingoutthephasesignal at the correct spatial frequency. Advantages of the YI method include thedetection of the whole interference pattern, contributing to the simplicity of thedevice.Furthermore,theidenticallengthsofthearmsreducetheeffectsofwavelengthdrift and temperature. Moreover, the YI biosensors are among the most sensitivebiosensors.AnLODof6x10‐8RIU,correspondingtoasurfacemassdensityof0.020pg/mm2,was reportedbyYmeti et. al [47]. Furthermore, Schmitt et. al reported anLODof9x10‐9RIU,correspondingtoasurfacemassdensityof0.013pg/mm2[48].Additionally,multiplexingcanbeachieved,byusingmorethantwowaveguides[49].The distance required from the output of the chip to the detector for detectinginterference and in order to get amaximum resolution can be a drawback for LOCplatforms.Moreover,thisdistanceLaswellasl,dandλshouldbekeptstabletoonlybesensitivetochangesinNeff.

Fig.1.11: Schematic overview of a Young interferometer, inwhich an analyte binding to anantibody causes a change in velocity in themeasuring arm compared to the reference arm,resultinginashiftoftheinterferencepattern.Thecoreofthewaveguideconsistsinthiscaseofthelayerabovethesubstrateandtheridges(whichprovidesconfinementinlateraldirection)ontopofthislayerwhicharemadefromthesamematerialandthereforecolouredthesame.

Introduction 17

Table1.1:ComparisonofdetectionlimitsofopticalbiosensorsDevice Surfacedetectionlimit RIdetectionlimit(RIU) References

SPR 1‐5pg/mm2(averaged)2fg/mm2(combinedwithimaging)

10‐5‐10‐7 [10,50]

Gratingcouplers 0.3pg/mm2 <10‐6 [19]Ringresonatorsensors 1.5pg/mm2 7.6x10‐7 [22,23]PCbasedsensor 2.1–490pg/mm2 2.4x10‐3‐7x10‐5 [25‐27,29]Interferometricwaveguidesensors

Mach‐Zehnderinterferometer

0.060‐1pg/mm2 1.9x10‐7‐5x10‐8 [32‐37]

Hartmaninterferometer n.d. 10‐6 [41]Bimodalwaveguide n.d. 2‐3x10‐7 [42‐44]Dualpolarizationinterferometer

<0.1pg/mm2 10‐7 [4,46]

Younginterferometer 0.013‐0.020pg/mm2 6x10‐8‐9x10‐9 [47,48]

1.4Methodsforimprovingspecificity

As discussed previously, the specificity is an important criterion of a biosensor.Despite the application of a bioreceptor layer and the techniques for improvingspecificityofabiosensor(referencechannels,blockingtechniquesandwashingsteps),thespecificityisofteninsufficient.Whenmeasuringahumansample,suchasbloodorserum,bulkeffectsandnon‐specificbindinglimitthespecificity.Therefore,Worthetal.developedamethodtoreducethecontributiontoRIchangesattributabletonon‐specific binding [51]. By tuning the evanescent field of two different polarizationmodes, a thin layer (20‐30nm) was desensitized and the response to non‐specificbindingwasreducedbyafactorofahundredormore.Theirpolarimetricwaveguideinterferometerusesthedifferentialphaseshiftbetweentwoorthogonalmodes,whichwere tuned by changing the core thickness such that both modes have equalsensitivity in the first20‐30nmof the evanescent field.Therefore, the contributionduetonon‐specificbindingwasreduced.However,anyRIchangeinthislayercannotbedetected,resultingalso inareductionof thecontributionduetospecificbinding.Furthermore, dual‐wavelength operation of an integrated‐optical differenceinterferometerwasusedtodiscriminatebetweenthebindingofmoleculesandbulkRIchanges,orbetweenbindingofmoleculesandtemperaturechanges[52].However,thevariousbackgroundcontributionstothesignalareusuallypresentsimultaneouslyandthereforetheexistingmethodsthatallowdistinguishingonlyoneof thevariousbackgroundcontributionsfromthesignalareinpracticenotalwayssufficient.

Inthisthesis,wepresentanewmethodtodiscriminatebetweenanalytesbasedon their size (size‐selective detection) which can be used to improve the sensorspecificity. We use the size‐selective detection in combination with the extremely

18 Chapter1

sensitiveYIbiosensor,butitisalsoapplicabletoothertypesofevanescentfield‐basedopticalsensors.Previously,weexpandedtheexistingdual‐wavelengthapproach[52]to a three‐wavelength approach that allows discrimination of several differentbackgroundcontributions(bulkandtemperatureinducedRIchanges)simultaneously[53].Here,weusemultiplewavelengths forsize‐selectivedetectionofanalytes.ThemultiplewavelengthsenabletheprobingofRIchangesatvariousdistancesfromthesensor surface, allowing the simultaneous discrimination of larger particles (e.g.viruses) from both smaller particles (e.g. proteins) and bulk contributions, whichimprovesthespecificityofthesensor.Themethodcanbeusedincombinationwithabioreceptor layer and the existing methods for improving the specificity of thebiosensors.

1.5Outlineofthethesis

InChapter2atheoreticalanalysisofsize‐selectivedetectionispresented.Numericalcalculations are used to optimize sensor design and the detection method. Next,Chapter3presentsthedesign,realisationandcharacterizationofthesetupwhich isused for size‐selective detection. Chapter 5 present different analysis approacheswhichcanbeused forsize‐selectivedetection.Subsequently, chapter4presents theapplicationoftheseanalysisapproachesonmeasurements,whereamongotherthingsbindingofbeads(size≈85nm,representingspecificbinding) isdiscriminatedfrombinding of protein A (size ≈ 2 nm, representing non‐specific binding) and bulkrefractiveindexchangesinducedbyD‐Glucose(whichoccursinthewholeevanescentfield of a few hundreds of nanometres which is used for detection). In order tooptimize the size‐selective detection method, the use of multiple wavelengths incombinationwithmultiplepolarizationswas investigated inChapter6.Moreover, itshows proof‐of‐principle experiments of the realization of size‐selective detectionbased on multiple wavelengths and polarizations. The measured data of chapter 5were analysed carefully and showed a timing difference betweenbulk changes andbindingofproteinswhichwasinvestigatedinChapter7.Finally,Chapter8presentsaforward look with possible new applications of size‐selective detection based onconstructivetechnologyassessment. Itwas investigated if themethodissuitable forbiosensing or if there are other more interesting or promising applications ormarkets.

Acknowledgements

Thiswork is supportedbyNanoNextNL, amicroandnanotechnology consortiumoftheGovernmentoftheNetherlandsand130partners.

Introduction 19

References

1. D.Duval,A.B.González‐Guerrero,S.Dante,C.Domínguez,andL.M.Lechuga,“InterferometricwaveguidebiosensorsbasedonSi‐technologyforpoint‐of‐carediagnostic,”inSiliconPhotonicsandPhotonicIntegratedCircuitsIII,2012),84310P‐84310P‐84311.

2. W.E.Moerner,“Newdirectionsinsingle‐moleculeimagingandanalysis,”Proc.Natl.Acad.Sci.U.S.A.104,12596‐12602(2007).

3. W.G.CoxandV.L.Singer,“FluorescentDNAhybridizationprobepreparationusingaminemodificationandreactivedyecoupling,”BioTechniques36,114‐122(2004).

4. M.C.Estevez,M.Alvarez,andL.M.Lechuga,“Integratedopticaldevicesforlab‐on‐a‐chipbiosensingapplications,”LaserPhoton.Rev.6,463‐487(2012).

5. S.Paulie,H.Perlmann,andP.Perlmann,“Enzyme‐linkedImmunosorbentAssay,”ineLS(JohnWiley&Sons,Ltd,Chichester,2001).

6. J.E.Butler,“Enzyme‐LinkedImmunosorbentAssay,”J.Immunoassay21,165‐209(2000).

7. K.B.MullisandF.A.Faloona,“SpecificsynthesisofDNAinvitroviaapolymerase‐catalyzedchainreaction,”MethodsEnzymol.155,335‐350(1987).

8. X.D.Fan,I.M.White,S.I.Shopova,H.Y.Zhu,J.D.Suter,andY.Z.Sun,“Sensitiveopticalbiosensorsforunlabeledtargets:Areview,”Anal.Chim.Acta620,8‐26(2008).

9. M.Piliarik,Homola,J.,“SPRSensorInstrumentation”inSurfacePlasmonResonanceBasedSensorsJ.Homola,ed.(SpringerBerlinHeidelberg,Berlin,2006),pp.95‐116.

10. J.Homola,“Surfaceplasmonresonancesensorsfordetectionofchemicalandbiologicalspecies,”ChemicalReviews108,462‐493(2008).

11. F.C.Chien,C.Y.Lin,J.N.Yih,K.L.Lee,C.W.Chang,P.K.Wei,C.C.Sun,andS.J.Chen,“Coupledwaveguide–surfaceplasmonresonancebiosensorwithsubwavelengthgrating,”Biosens.Bioelectron.22,2737‐2742(2007).

12. J.Dostálek,J.Homola,andM.Miler,“Richinformationformatsurfaceplasmonresonancebiosensorbasedonarrayofdiffractiongratings,”Sens.Actuators,B107,154‐161(2005).

13. J.Dostálek,J.Čtyroký,J.Homola,E.Brynda,M.Skalský,P.Nekvindová,J.Špirková,J.Škvor,andJ.Schröfel,“Surfaceplasmonresonancebiosensorbasedonintegratedopticalwaveguide,”Sens.Actuators,B76,8‐12(2001).

14. A.Suzuki,J.Kondoh,Y.Matsui,S.Shiokawa,andK.Suzuki,“Developmentofnovelopticalwaveguidesurfaceplasmonresonance(SPR)sensorwithduallightemittingdiodes,”Sens.Actuators,B106,383‐387(2005).

15. R.SlavíkandJ.Homola,“Ultrahighresolutionlongrangesurfaceplasmon‐basedsensor,”Sens.Actuators,B123,10‐12(2007).

16. “BiacoreTM”(GeneralElectricCompany),retrieved2013http://www.biacore.com/lifesciences/index.html

17. W.Lukosz,D.Clerc,P.M.Nellen,C.Stamm,andP.Weiss,“Outputgratingcouplersonplanaropticalwaveguidesasdirectimmunosensors,”Biosens.Bioelectron.6,227‐232(1991).

18. A.Brandenburg,R.Polzius,F.Bier,U.Bilitewski,andE.Wagner,“Directobservationofaffinityreactionsbyreflected‐modeoperationofintegratedopticalgratingcoupler,”Sens.Actuators,B30,55‐59(1996).

19. K.Cottier,M.Wiki,G.Voirin,H.Gao,andR.E.Kunz,“Label‐freehighlysensitivedetectionof(small)moleculesbywavelengthinterrogationofintegratedopticalchips,”Sens.Actuators,B91,241‐251(2003).

20. M.WikiandR.E.Kunz,“Wavelength‐interrogatedopticalsensorforbiochemicalapplications,”Opt.Lett.25,463‐465(2000).

20 Chapter1

21. “MicroVacuumLtd.”,retrieved2013,http://www.owls‐sensors.com/company.aspx.22. M.Iqbal,M.A.Gleeson,B.Spaugh,F.Tybor,W.G.Gunn,M.Hochberg,T.Baehr‐Jones,R.

C.Bailey,andL.C.Gunn,“Label‐FreeBiosensorArraysBasedonSiliconRingResonatorsandHigh‐SpeedOpticalScanningInstrumentation,”IEEEJ.Sel.Top.Quant.Electron.16,654‐661(2010).

23. M.S.Luchansky,A.L.Washburn,T.A.Martin,M.Iqbal,L.C.Gunn,andR.C.Bailey,“Characterizationoftheevanescentfieldprofileandboundmasssensitivityofalabel‐freesiliconphotonicmicroringresonatorbiosensingplatform,”Biosens.Bioelectron.26,1283‐1291(2010).

24. “Genalyte”,retrieved2015,http://www.genalyte.com/about‐us/our‐technology/.25. S.MandalandD.Erickson,“Nanoscaleoptofluidicsensorarrays,”Opt.Express16,

1623‐1631(2008).26. D.Dorfner,T.Zabel,T.Hürlimann,N.Hauke,L.Frandsen,U.Rant,G.Abstreiter,andJ.

Finley,“Photoniccrystalnanostructuresforopticalbiosensingapplications,”Biosens.Bioelectron.24,3688‐3692(2009).

27. J.García‐Rupérez,V.Toccafondo,M.J.Bañuls,J.G.Castelló,A.Griol,S.Peransi‐Llopis,andÁ.Maquieira,“Label‐freeantibodydetectionusingbandedgefringesinSOIplanarphotoniccrystalwaveguidesintheslow‐lightregime,”Opt.Express18,24276‐24286(2010).

28. Y.LiuandH.W.M.Salemink,“Photoniccrystal‐basedall‐opticalon‐chipsensor,”Opt.Express20,19912‐19920(2012).

29. S.Kita,K.Nozaki,andT.Baba,“Refractiveindexsensingutilizingacwphotoniccrystalnanolaseranditsarrayconfiguration,”Opt.Express16,8174‐8180(2008).

30. S.C.Buswell,V.A.Wright,J.M.Buriak,V.Van,andS.Evoy,“Specificdetectionofproteinsusingphotoniccrystalwaveguides,”Opt.Express16,15949‐15957(2008).

31. P.Kozma,F.Kehl,E.Ehrentreich‐Förster,C.Stamm,andF.F.Bier,“Integratedplanaropticalwaveguideinterferometerbiosensors:Acomparativereview,”Biosens.Bioelectron.58,287‐307(2014).

32. R.G.HeidemanandP.V.Lambeck,“Remoteopto‐chemicalsensingwithextremesensitivity:design,fabricationandperformanceofapigtailedintegratedopticalphase‐modulatedMach–Zehnderinterferometersystem,”Sens.Actuators,B61,100‐127(1999).

33. S.Dante,D.Duval,B.Sepúlveda,A.BelenGonzález‐Guerrero,J.RamónSendra,andL.M.Lechuga,“All‐opticalphasemodulationforintegratedinterferometricbiosensors,”Opt.Express20,7195‐7205(2012).

34. K.Zinoviev,L.G.Carrascosa,J.SánchezdelRío,B.Sepúlveda,C.Domínguez,andL.M.Lechuga,“SiliconPhotonicBiosensorsforLab‐on‐a‐ChipApplications,”Adv.Opt.Tech.2008,6(2008).

35. P.Kozma,A.Hámori,S.Kurunczi,K.Cottier,andR.Horvath,“Gratingcoupledopticalwaveguideinterferometerforlabel‐freebiosensing,”Sens.Actuators,B155,446‐450(2011).

36. D.Duval,J.Osmond,S.Dante,C.Dominguez,andL.M.Lechuga,“GratingcouplersintegratedonMach‐Zehnderinterferometricbiosensorsoperatinginthevisiblerange,”IEEEPhotonicsJ.5,3700108‐3700108(2013).

37. P.V.Lambeck,“Integratedopticalsensorsforthechemicaldomain,”Meas.Sci.Technol.17,R93‐R116(2006).

38. “Optisense”,retrieved2015,http://www.optisense.nl/component/option,com_frontpage/Itemid,1/.

39. “CreoptixGmbH”,retrieved2015,http://www.creoptix.com/index.html.40. B.H.Schneider,J.G.Edwards,andN.F.Hartman,“Hartmaninterferometer:versatile

integratedopticsensorforlabel‐free,real‐timequantificationofnucleicacids,proteins,andpathogens,”Clin.Chem.43,1757‐1763(1997).

Introduction 21

41. J.Xu,D.Suarez,andD.S.Gottfried,“Detectionofavianinfluenzavirususinganinterferometricbiosensor,”Anal.Bioanal.Chem.389,1193‐1199(2007).

42. D.Duval,A.B.Gonzalez‐Guerrero,S.Dante,J.Osmond,R.Monge,L.J.Fernandez,K.E.Zinoviev,C.Dominguez,andL.M.Lechuga,“Nanophotoniclab‐on‐a‐chipplatformsincludingnovelbimodalinterferometers,microfluidicsandgratingcouplers,”LabChip12,1987‐1994(2012).

43. K.E.Zinoviev,A.B.González‐Guerrero,C.Domínguez,andL.M.Lechuga,“IntegratedBimodalWaveguideInterferometricBiosensorforLabel‐FreeAnalysis,”J.LightwaveTechnol.29,1926‐1930(2011).

44. S.Dante,D.Duval,D.Fariña,A.B.González‐Guerrero,andL.M.Lechuga,“Linearreadoutofintegratedinterferometricbiosensorsusingaperiodicwavelengthmodulation,”LaserPhoton.Rev.9,248‐255(2015).

45. G.H.Cross,A.A.Reeves,S.Brand,J.F.Popplewell,L.L.Peel,M.J.Swann,andN.J.Freeman,“Anewquantitativeopticalbiosensorforproteincharacterisation,”Biosens.Bioelectron.19,383‐390(2003).

46. “Farfield”(BiolinScientific),retrieved2015,http://www.farfield‐group.com/technology_dpi.asp.

47. A.Ymeti,J.S.Kanger,J.Greve,G.A.J.Besselink,P.V.Lambeck,R.Wijn,andR.G.Heideman,“Integrationofmicrofluidicswithafour‐channelintegratedopticalYounginterferometerimmunosensor,”Biosens.Bioelectron.20,1417‐1421(2005).

48. K.Schmitt,B.Schirmer,C.Hoffmann,A.Brandenburg,andP.Meyrueis,“Interferometricbiosensorbasedonplanaropticalwaveguidesensorchipsforlabel‐freedetectionofsurfaceboundbioreactions,”Biosens.Bioelectron.22,2591‐2597(2007).

49. A.Ymeti,J.S.Kanger,J.Greve,P.V.Lambeck,R.Wijn,andR.G.Heideman,“RealizationofamultichannelintegratedYounginterferometerchemicalsensor,”Appl.Opt.42,5649‐5660(2003).

50. S.P.Wang,X.N.Shan,U.Patel,X.P.Huang,J.Lu,J.H.Li,andN.J.Tao,“Label‐freeimaging,detection,andmassmeasurementofsinglevirusesbysurfaceplasmonresonance,”Proc.Natl.Acad.Sci.U.S.A.107,16028‐16032(2010).

51. C.Worth,B.B.Goldberg,M.Ruane,andM.S.Unlu,“Surfacedesensitizationofpolarimetricwaveguideinterferometers,”IEEEJ.Sel.Top.Quant.Electron.7,874‐877(2001).

52. C.Stamm,R.Dangel,andW.Lukosz,“Biosensingwiththeintegrated‐opticaldifferenceinterferometer:dual‐wavelengthoperation,”Opt.Commun.153,347‐359(1998).

53. A.Ymeti,“DevelopmentofamultichannelintegratedYounginterferometerimmunosensor,”UniversityofTwente(2004).

Chapter21

Size‐selectivedetectioninintegratedopticalinterferometricbiosensors

AbstractIn this chapter we present a size‐selective detection method for integrated opticalinterferometric biosensors that could strongly enhance their performance. Wedemonstrate that by launching multiple wavelengths into a Young interferometerwaveguidesensoritisfeasibletoderiverefractiveindexchanges( n )fromdifferentlayers above the waveguide surface, enabling one to distinguish between boundparticles (e.g. proteins, viruses, bacteria) based on their differences in size andsimultaneously eliminating interference from a n in the bulk (region of a fewhundred nanometres above sensor surface). Numerical calculations are used tooptimizesensordesignandthedetectionmethod.Addingsize‐selectivitytothesensorreducesthesensitivityofthesensor.However,thetheoreticalsensitivityremainsstillcomparabletootherexistingbiosensorswhendiscriminatingbetween n ’sinthreedifferent layers above the waveguide based on simultaneous detection of effectiverefractiveindexchanges( effN )atthreedifferentwavelengths.Assumingaparticleof

80nminsizeasthespecificanalytetodetect,thetheoreticallydeterminedminimumdetectablemasscoverage is4×102 fg/mm2(assumingaphasenoiseof10‐4 fringes).This is approximately one order ofmagnitude higher than theminimumdetectablemass coverage with the YI using a single wavelength. However, with size‐selectivedetectionitisnowpossibletodiscriminatethe80nmsizedanalytebindingfromnon‐specificboundparticlesof10nmsizeandsimultaneouslyoccurringbulkchanges.

Partofthischapterwaspublishedin:H.K.P.Mulder,A.Ymeti,V.Subramaniam,J.S.Kanger,“Size‐SelectiveDetectioninIntegratedOpticalInterferometricBiosensors”,OpticsExpress,20(19),20934‐20950,2012

24 Chapter2

2.1Introduction

Integrated optical (IO) biosensors have been demonstrated as a powerful detectionand analysis tool for biosensing. Main advantages of IO biosensors are its highsensitivity, real‐time and label‐free measurements. Interferometric sensors [1‐6],grating couplers [7, 8], resonant optical microcavity sensors [9‐11], and photoniccrystalwaveguidesensors[12,13],areseveralIOsensorswhichhavebeendeveloped.Integrated optical interferometric biosensors sense refractive index (RI) changes,inducedbyanalytebinding,occurringintheevanescentfield.Thesesensors,includingtheMachZehnderinterferometerandtheYounginterferometer(YI),showextremelyhigh(10‐7‐10‐8refractiveindexunits(RIU))RIsensitivity.TheYIisastrongcandidateforpoint‐of‐careviraldiagnostics,becauseofthishighsensitivityanditsmultiplexingcapability [14].Measurementsshowshort timedelays,becausenoextensivesampletreatmentisneeded.However,theutilizationofthehighsensitivityisoftenhamperedby background signals arising from non‐specific RI changes within the evanescentfield.AnyRIchangewithintheevanescentfieldwillcontributetothemeasuredsignal.Consequently,inadditiontospecificbindingoftheanalyte,alsonon‐specificbindingandRIchanges(e.g.duetotemperaturechanges)inthefluidcoveringthewaveguide(bulk) will be detected. To distinguish between specific and non‐specific binding,selective chemical binding techniques are used in combination with washing stepsand/ordifferentialmeasurements.Nevertheless,non‐specificityandbulkbackgroundchangesstillhampersuccessfulapplicationofthesetypeofbiosensors.Measurementsdone inbody fluids suchasbloodserumshowahighvariability inbackgroundandlargenon‐specificbinding.Forthatreason,amethodtoreducethecontributiontoRIchanges attributable to non‐specific binding was developed [15]. By tuning theevanescent field of two different polarization modes a thin layer (20‐30nm) wasdesensitized and the response to non‐specific binding was reduced by a factor ofhundredormore.Furthermore,adual‐wavelengthoperationofanintegrated‐opticaldifferenceinterferometerwasusedtodiscriminatebetweenbindingofmoleculesandbulk RI changes or between binding of molecules and temperature changes [3]. Ingeneral however, the various background contributions to the signal are presentsimultaneouslyandthereforetheexistingmethodsthatallowdistinguishingonlyoneofthemfromthesignalareinpracticenotalwayssufficient.Previouslyweexpandedtheexistingdual‐wavelengthapproach[3]toathreewavelengthapproachthatallowsto discriminate several different background (bulk and temperature induced RIchanges)contributionssimultaneously [16].Hereweexplorea similarapproach forsize‐selective detection of analytes. The use of multiple wavelengths (3 or more)enablestoprobeRIchangesatdifferentdistancesfromthesensorsurfaceallowingtodiscriminate largerparticles(e.g.viruses) frombothsmallerparticles(e.g.proteins)andbulkcontributions.Weprovidea theoreticalbasis for thismethod,weoptimize

Size‐selectivedetectioninintegratedopticalinterferometricbiosensors 25

themethod forapplication toaYI sensorandwecalculate the achievabledetectionlimit. We anticipate that using the size‐selective multiple‐wavelength approach aspresentedhereshouldsignificantlyimprovethebackgroundsuppression.Itshouldbenoted that themethodpresentedherewillmost likelynot replaceexistingmethodslikebio‐receptorlayersandantifoulingstrategiestoreducenon‐specificbinding[17],but is rather to be used in combination with these methods to yield enhancedspecificity.

Adetailedtheoreticalanalysisisgivenontheperformanceofthisnewmethodfortwocases.Thefirstcaseisaimedtodistinguishbetweenspecificbindingofanalytes(~80nm,e.g.virusparticles)andbulkchangesbyusingtwowavelengths.Thesecondcaseuses threewavelengths todistinguishspecificbindingof largerparticles (~80nm) frombulk contributions andnon‐specific binding (smaller particles 10nm, e.g.proteins).Optimizedwaveguidestructuresarecalculatedforeachcase.AlthoughwespecificallydevelopthismethodforaYIsensor,themethodisalsoapplicabletoothertypesofIOinterferometricsensors.

2.2Theoreticalaspects

ThissectionstartswiththetheoryrequiredtocalculatetheprecisionwithwhichtheRI change can be determined from the phase changes measured for the differentwavelengths. This approach is used to optimizewaveguide properties. It should benoted that the precision (defined as the standard deviation n of subsequent

measurements of the RI change) is the relevant parameter for indicating theperformance of the sensor. Induced RI change due to specific binding of analytesshouldexceed2× n (95%confidenceinterval).

2.2.1GeneraltheoryYIsensorsarebasedontheevanescentfieldsensitivityofguidedmodespropagatingthroughthewaveguidestructureofthesensor[18].Fig.2.1illustratestheworkingoftheYIsensor.Monochromaticlightiscoupledintoanopticalchannelwaveguideandsplit into two channels, including ameasurement and a reference channel. BindingeventsnearthesurfaceofthemeasurementchannelresultinanRIchangenatthissurface. Consequently, the phase of the beam in the measuring channel changes,resulting in an alteration of the interference pattern that exist in the region ofoverlappingbeamsfromthe twochannels.AssumingsmallRIchangessuchthattheelectricfielddistributionoftheguidedmode(modeprofile)isnotaffected,thephasechange between two beams, propagating through any two channels, can bedescribedby[19]:

26 Chapter2

2 2 eff

eff

Nl N l n

n

(2.1)

where l is the length of the sensing window, is the vacuumwavelength of theguidedlight, effN theeffectiveRIchangeoftheguidedmode,n theRIchangeinthe

region probed by the evanescent field and effN n

the sensitivity coefficient of

effN withrespectton,forawavelength .Althoughnotexplicitlywritten,chromaticdispersion is taken intoaccount (seeAppendix2.A).Next,wedefinemultiple layersabove the coreof thewaveguideofwhich theRI changehas tobedetermined (Fig.2.2). Thicknesses of the defined layers can be chosen arbitrarily depending on theexperiment, e.g. three layers to discriminate between non‐specific binding (e.g.proteins),specificbinding(e.g.viruses)andbulkRIchanges(seeFig.2.3).

Fig.2.1:TheprincipleofaYounginterferometer,wherelightiscoupledinandguidedthroughan integratedchannelwaveguide structureandprojectedontoaCCDcamerabya cylindricallens,givinganinterferencepattern.Thecoreofthewaveguideconsistsinthiscaseofthelayerabovethesubstrateandtheridges(whichprovidesconfinementinlateraldirection)ontopofthislayerwhicharemadefromthesamematerialandthereforecolouredthesame.


Fig.2.2: Structure definition ofwaveguidewith on topN introduced imaginary layers and aguidedmodeprofile(dashedline),wheredisthethicknessandntherefractiveindex.

Fig.2.3:Guidedmodeprofilesofthreedifferentwavelengthspropagatingthroughawaveguidestructurewiththreelayersintroducedontopofthesensingwindowtodistinguishbetweenthenon‐specificproteinbindingof smallermolecules, thespecificbindingof largerparticles,andthebulksolutionchanges.

28 Chapter2

Theelectric fielddistributionof theguidedmodedependson thewavelengthofthelight(shorterwavelengthsaremoreconfinedtothecorethanlongerwavelengths,seeFig.2.3).Consequently,theRIchangesinthedifferentlayerscanbedeterminedbymeasuring the phase changes at a number of different wavelengths, provided thenumberoflayersdoesnotexceedthenumberofusedwavelengths.

Consider layerN layers(seeFig.2.2),and N numberofdifferentwavelengths.The

measuredphasechangescanbewrittenas(inanalogywithEq.(2.1)):

11

,

layer

ijs

NN

n

nM n with n

n

(2.2)

with j the phase change measured at j , in the RI change in layer i, and ..

(sensitivitymatrix)definedas:

1,1 2,1 3,1 ,11 1 1 1

1,2 ,22 2

1,3 , ,33 3

1, 2, 3, ,

1 1 1 1

1 1

1 1 12

1 1 1 1

layer

layer

layer

layer

N

N

si j N

j

N N N N NN N N N

S S S S

S S

M l S S S

S S S S

(2.3)

with ,j

i j eff iS N n

thesensitivitycoefficientofthe thi layerand thj wavelength

(seeAppendix2.Bforanexplicitexpressionof ,i jS ).Eq.(2.2)isrewrittentofindthe

RIchangeineachlayer:

1

sn M

, (2.4)

where 1sM istheinverseof sM forasquarematrixandtherightinverseof sM fora

non‐squarematrix (for layerN N ). Eq. (4) has a unique solution if det( ) 0sM . In

thatcase,theRIchangeinlayericanbedeterminedwithaprecisionin

(definedas

thestandarddeviationin in )dependingontheprecisioni

ofthemeasurement


of j ,whichisdeterminedbyexperimentalfactorssuchaslasernoise,cameranoise,

andtemperaturefluctuations.Ifthematrix sM getsmoresingulartheprecisionin

willworsen.Therefore,itisessentialtooptimizetheexperimentalconfigurationsuch

that sM doesnotgetsingular. sM isdeterminedbythesensitivitycoefficientswhich

in turn depend on the guided mode profiles. If the mode profiles are similar, the

sensitivitycoefficientswillbealmostequalandasaconsequence,thematrix sM gets

more singular. Therefore, thewavelengths should differ asmuch as possible in theworkablewavelengthrange.

We define the relative precision (describing the relative precision in

determining the RI change in the thi layer) as the ratio betweenin

andi

.

Assuming 1 k

k N ,therelativeprecisionisavector withthe thi

element givenby(seeAppendix2.C):

1 2

21,

1

i

Nn

i ijj

Φ M

s . (2.5)

is evaluated for different wavelengths, waveguide refractive indices and layer

thicknesses. Given an experimentally determined precision in the phasemeasurements ( ≈ 10‐4 fringes @ 1 Hz for the reported YI sensor [18]), the

precisionin

with which a RI change of a given layer i can be measured, can be

calculatedbymultiplyingtherelativeprecision i ofthecorrespondinglayerwiththe

experimentallydeterminedvalueof .

2.2.2SurfacemasscoverageNext,weconverttheRIchangesobtainedfromameasurementintothemorerelevant

mass coverage of the specific bound particles sC . To discriminate between non‐

specific binding of smaller molecules (e.g. proteins), specific binding of largermolecules(e.g.viruses)andbulksolutionchangesweintroducethreelayersontopofourwaveguide(seeFig.2.3).Weassumethatlayer2changesduetospecificbindingandbulkRIchanges,whereaschangesinlayer3areonlycausedbybulkRIchanges.Ittherefore follows that the RI changes due to specific binding are given by thedifference in RI changes between layer 2 and 3. Multiplying this RI change with aconstantconvertsanRIchangeintoamasscoverage(seeAppendix2.D):

30 Chapter2

2 3

1( )sC n n

, (2.6)

andtheminimaldetectablemasscoveragegivenby:

2

viruss nC (2.7)

whereisgivenby:

2analyte solutionn n d m . (2.8)

Assuming that the specific boundparticles is a virus, theRI of the analyte is analyten

=1.41[20]andtheRIofabuffersolution solutionn =1.33,amolecularweightofasingle

Adenovirusparticle, analytem =1.75×108Da[21](=2.91×10‐1fg),andthediameterof

thespecificboundvirusparticle analyted =80nm, isequalto1.76×10‐9mm2/fg.

2.3Resultsanddiscussion

2.3.1.OptimizationofthewaveguidestructureFirst, the two layers case is treated,which is used to discriminate between analytebinding(layer1)andRIchangesof thebulk(layer2).Layer1, the “specificbindinglayer”hasadefinedthicknessof80nm.Thesecondlayeriscalled“bulk”andisallthespaceabove layer1.Thedependenceof the relativeprecisionon thecore thicknessdcore is given (Fig. 2.4a,b) for three different core refractive indices. The refractiveindicesarechosensuchthattheycorrespondtothoseofrealmaterialsthatareoftenused forwaveguide fabrication;n≈1.77: aluminiumoxide (Al2O3), n≈2.02: siliconnitride(Si3N4)andn≈2.65titaniumoxide(TiO2).Theusedwavelengthsare 1 =400

nmand 2 =700nm.MoredetailsofthewaveguidecanbefoundinTable2.1.Forall

simulationsonlyzerothordertransverseelectric(TE0)modesaretakenintoaccount.

FromFig.2.4itisclearthatthebestperformance(lowestvalueof )isobtainedforhigh values of ncore combined with low core thicknesses. This holds for both theprecision achieved for the specific binding layer and the bulk layer (note that atoptimal thickness, the dependence on the core RI is marginal). This conclusion isimportantinchoosingthewaveguidematerials,i.e.choosingSi3N4(n( =500nm)=2.034)asacorematerialisjustifiedasitcombinesahighRIwithexcellentcleanroom

processingpossibilities.Theinterpretationofthecalculated anditsconsequencesforthemasscoveragearediscussedlater.


Fig.2.4:Relativeprecisionasafunctionofthecorethicknessandcorerefractiveindex(givenatawavelengthof550nm)fora)thespecificlayerusingtwolayers,b)thebulkusingtwolayers,c) thespecificbinding layerusingthree layers,d) thebulkusingthree layersande) thenon‐specificbindinglayerusingthreelayers.

40 80 120 160 200

0

1

2

3

4

40 80 120 160 2000

2

4

6

8

10

12

14

40 80 120 160 200

0

1

2

3

4

5

6

40 80 120 160 200

0

2

4

6

8

10

40 80 120 160 200

0

1

2

3

n=1.77(Al2O3)

n=2.02(Si3N4)

n=2.65(TiO2)

x10‐3

Specificbindinglayer

(r

ad-1)

dcore(nm)

Specificbindinglayer

e

db

c x10‐3

n=1.77(Al2O3)

n=2.02(Si3N4)

n=2.65(TiO2)

(r

ad-1)

dcore(nm)

a

x10‐2 Bulk

n=1.77(Al2O3)

n=2.02(Si3N4)

n=2.65(TiO2)

(r

ad-1)

dcore(nm)

x10‐2 Bulkn=1.77(Al

2O3)

n=2.02(Si3N4)

n=2.65(TiO2)

(r

ad-1)

dcore(nm)

x10‐2 Non‐specificbindinglayern=1.77(Al

2O3)

n=2.02(Si3N4)

n=2.65(TiO2)

(r

ad-1)

dcore(nm)

32 Chapter2

Table2.1:Waveguide structure details for all the simulations, where the RI of the layers isdeterminedbytheSellmeierequations(seeAppendixA).Simulationdescription

Substratematerial

Corematerial1 1Nn n cored

(nm)1 1Nd d

(nm)1d

(nm)2d

(nm)

vs. cored

(twolayers)

SiO2(n=1.461@=550nm),[21]

Al2O3,(n=1.770@=550nm),[22]Si3N4,(n=2.024@=550nm),[23]TiO2,(n=2.648@=550nm),[22]

Water@20°C,(n=1.335@=550nm),[24]

30‐200

‐ 80 ‐

vs. cored

(threelayers)

SiO2 Al2O3,Si3N4,TiO2 Water@20°C

30‐200

‐ 10 80

vs. layerN SiO2 Si3N4 Water@20°C

70 randomvalues

‐ ‐

vs. N SiO2 Si3N4 Water@

20°C70 ‐ 80 ‐

Inordertodistinguishbetweenspecificanalytebinding,non‐specificbindingand

bulk RI changes, three layers are required, a non‐specific binding layer, a specificbindinglayer,andabulklayer.FordetailsseeTable2.1.Fig.2.4c‐eshowsΦ forthespecific binding layer, the bulk and the non‐specific binding layer respectively as a

functionof cored forthedifferentcorematerials.Thethreewavelengthsare: 1 =400nm, 2 =550nmand 3 =700nm.Foroptimaldetectionof specificbinding, the

relative precision of the specific binding layer and the bulk should beminimal. Fig.

2.4cshowsaminimumvalueof specificΦ =6.12×10‐4rad‐1 foracorematerialofTiO2

and cored =35nm,andatwo‐foldhighervalueof specificΦ =1.63×10‐3rad‐1forSi3N4as

corematerialand cored =70nm.Fig.2.4dshowsasimilarrelativeprecisionofthebulk

for both conditions. For each core material, the core thickness should be chosencarefully.

2.3.2.ExpandingthenumberoflayersWith the use of an increasing number of wavelengths it is, in theory, possible todetermineacompleteRIchangeprofile,i.e.theRIchangeasafunctionofdistancetothe waveguide surface. However the use of an increasing number of wavelengthsresults in mode profiles which are increasingly similar and therefore result in a

worseningprecisionin

. Forexample,bycomparing the relativeprecisions for the

twoand three layerscases,weseean increase in ofapproximatelyoneorderof

magnitude. In order to explore the limits,we calculated the as a functionof the


number of layers layerN . TheRI change cannot be determined at distances from the

waveguide surface where the evanescent field becomes very small. Therefore, wedefineourlayersintheregionstartingatthesurfaceandendingat200nmfromthesurface.Thespaceabove the200nm limit isconsideredasbulk.The thicknessesof

the layers are randomly chosenwith aminimum of 10 nm. is calculated for alllayersandthiscalculationisrepeated1000times(eachtimewithdifferentrandomly

chosenlayerthicknesses).Next,themeanofallthesecalculatedrelativeprecisions is calculated.Thenumber ofwavelengths is chosenequal to layerN . For the corewe

chose Si3N4 with a thickness of 70 nm. For optimal settings, the wavelengths arespread maximally. This means that min = 400 nm and max = 700 nm. The

wavelengthsinbetweenareequallydividedwithinthisrange.Fig.2.5shows asa

functionof layerN ,wherebulk is included in layerN ,butnot in .Theerrorbarsare

basedona95%confidenceintervalofthe1000calculatedrelativeprecisionsandgivean indication about the spread in the calculated relative precisions. A one‐layersituationis includedasasolid lineinFig.2.5(wavelengthissetat550nm).Fig.2.5

showsthat increasesexponentiallywith,meaningthat canbedeterminedlessprecisely. For a situation with three layers and assuming a phase precision ofapproximately10‐4fringesfortheYIsensor[18],thedetectionlimit(whichisdefinedhereas theminimumreliablydetectable inducedRIchangewhichequals twotimesthe precision) of n is on average 5x10‐6 RIU, which is still comparable to otheropticalbiosensors[25].

Fig.2.5:Mean relativeprecisionas a functionof thenumberof layers,which is equal to thenumber ofwavelengths, for a core thickness of 70 nm and Si3N4 as corematerial. Themeanrelative precision multiplied by the phase precision gives the minimal reliably detectableinduced n .

2 3 4 5 6 7 810

‐5

10‐3

10‐1

101

103

105

single

proteinmonolayer

<>(r

ad-1)

Nlayer

34 Chapter2

2.3.3.ExpandingthenumberofwavelengthsThenumberofwavelengths shouldbeequal toor larger than thenumberof layersthatneedstoberesolved.Intheprevioussectionsweexploredseveralconfigurations

ofvariousnumberof layersandequalnumberofwavelengths.Here isevaluatedfor the case layerN = 2 and layerN N . The two layers are defined as a “specific

binding layer”of80nmanda “bulk” layercoveringall thespaceabove80nm.Thewavelengths are in each case equally divided between 400 nm and 700 nm (for

waveguide details see Table 2.1). Fig. 2.6 shows as a function of the number ofwavelengths for layerN = 2. The for the “specific binding layer” only decreases a

factor of ≈ 2 by adding eight extrawavelengths,where of bulk layer decreases

even less. It can be concluded that decreases with an increasing number ofwavelengths. In conclusion, adding more wavelengths than layers results in a

marginalgainin .

Fig.2.6:Relativeprecisionasa functionof thenumberofwavelengths for two layers,a corethicknessof70nmandSiO2asthesubstratematerial,Si3N4asthecorematerial,andwaterontopofthiscore.

2.3.4.SurfacemasscoverageTo illustratethepossibilitiesof thesize‐selectivedetectionscheme,wecalculate thesurface mass coverage for the specific case of virus detection. Using Eq. (2.7) incombination with the results from the three layer case, we calculate the minimal

detectable surfacemass coverage sC .Multiplying the relative precisions fromFig.

2.4c for the specific binding layer with the previously reported phase precision ofapproximately10‐4fringesfortheYIsensor[18]gives

analyten .Foracorethicknessof

70nmandacorematerialofSi3N4andassumingthatthespecificanalyteisavirus,we

2 4 6 8 100

4

8

12

16

20

24x10‐4

BulkSpecificbindinglayer

(rad

-1)

N


find sC =8×102 fg/mm2. Itshouldbenotedthatdifferentanalyteshavedifferent

sizes.InFig.2.4weanalysedaspecificsituationofan80nmsizedanalyte.However,calculations show that the values of the relative precision for analytes of differentsizes do not differ significantly, e.g. for analytes of 160 nm the relative precisiondiffersafactorof2.

2.3.5.DetectionofanalytesincomplexmatricesHerewewouldliketodiscussseveralissuesrelatedtotheuseofthismethodforthedetectionofanalytes incomplexmatricessuchasblood, serum,urineorsputum. Inouranalysisweassumedathinanduniformlayerofparticlesthatarenon‐specificallybound to the sensor surface. However, the particles responsible for non‐specificbindingcanhavevarioussizes,withsomeofthembeinglargerthan10nm,exceedingthethicknessofthenon‐specificbindinglayerassumedinthecalculations.Ingeneral,themajorityofparticles in e.g. bloodor serumarehowevermuch smaller than theanalyte,i.e.avirusparticle.Itisthismajority,theexcessofsmallerparticlescomparedtothenumberofanalytemoleculesthatleadstomeasurablebackgroundduetonon‐specific binding. Small numbers of very large particles next to the analytes can bediscriminatedbythespecificityofthebio‐receptorcoatingofthesensorsurface.

Anotherconcernisthatseveraltypesofparticlescanbepresentinarealsampleall having the same size. The method presented here obviously cannot distinguishbetweenparticleswiththesamesizeandthereforeitisimportanttoalwayscombinethismethodwith a traditional affinity bio‐receptor coating of the sensor surface inorder to strongly enhance the binding of the specific analyte of interest. Only inextremecaseswheretheconcentrationoftheanalyteofinterestismuchlowerthantheconcentrationofsimilarsizedparticlescouldthisresultinthenon‐specificbindingofothertheseparticlesexceedingthespecificbindingoftheanalyteofinterest.Inthiscase our method obviously cannot distinguish between the analytes and the non‐specificboundsimilarsizedparticles.

Finally, in a real sample the molecules are moving rather than being static.Brownianmotionofmoleculesinthesamplingvolumegivesrisetofluctuationsintherefractiveindex.Theexpectedfluctuationsintherefractiveindexina10nmthinlayerabovethesensorsurfaceiscalculatedtobe~10‐9RIU(correspondingtoaphasenoiseof~10‐6 fringes) for an average protein content of 10 g/ml (typical for blood). Thecalculatedfluctuationsarewellbelowthedetectionlimitofthesensor(twoordersofmagnitude smaller than the value used in the calculations) and as such we canconclude that fluctuations in the signal due to Brownianmotion of proteins in thesamplefluidcanbeneglected.

36 Chapter2

2.3.6.ImplementationFortheimplementationofthisnewmethodtwoaspectsneedspecialattention.First,multiple lasers emitting at different wavelengths should be coupled into a singlechannel waveguide of the YI sensor. This can be accomplished by combining theoutput of multiple monochromatic lasers using either free space optics (e.g. usingappropriatedichroicmirrors)orbyusingfiber‐opticalcombiners.Secondly,foreachwavelengththephasechangeshouldbemeasured independently fromtherecordedinterferencepattern.Thiscanbedoneintwoways.First,theinterferencepatternofeachwavelengthismeasuredseparately.Thiscanbedoneusingeitheramulticolourvideocamera,orbyintroducingadispersiveelement(e.g.agrating)inthedetectionpath to separate the different interference patterns (one for each wavelength)spatiallyonamonochromevideocamera.Alternatively, thesamedetectionsetupasusedforsinglewavelengthsensorscanbeused[1].Inthiscaseasingleinterferencepatternisrecorded.Toobtainthephasechangesforeachwavelengthseparatelyonecanmakeuseofthefactthatthespatialfrequencyoftheinterferencepatterndependsonthewavelengthoftheusedlight.Therefore,theamplitudespectrumoftheFourier‐transformed interference pattern now consists of well‐separated spatial frequencypeaks (one for each wavelength). Consequently, the phase change for a givenwavelengthcanbemonitoredindependentlyfromtheotherwavelengthsbyselectingthecorrespondingspatialfrequencyinthephasespectrumoftheFourier‐transformedinterferencepattern.Wehavesuccessfullyusedasimilarapproachtosimultaneouslymeasuremultiplephasechangesfromamulti‐channelintegratedopticalYIsensor[6].

2.3.7GeneraldiscussionComparing the calculated detection limit (which is defined here as the minimumreliably detectable induced RI change which equals two times the precision) withexisting methods we find that our method is comparable to the detection limit ofopticalbiosensorsbasedonSPR(10‐6RIU)[26].Thedetectionlimitalsocomparesto reported results obtained with grating couplers ( 10‐6 RIU, 0.3 pg/mm2) [8],photoniccrystals(10‐2RIU)[13]andresonantopticalmicrocavitiessensors(10‐5‐10‐6RIU)[10,11].AlthoughinthismethodthedetectionlimitgetsworsecomparedtosinglewavelengthYIsensors,theperformanceiscomparabletoexistingmethodswiththeadvantagethatthesensoriscapableofsizeselectivedetectionwhichwebelieveyieldsastrongimprovementinthespecificityofthesensor.

We further envision thepotential of implementing improvements or alternativeconfigurationstoimprovesensorperformance.OnepossibleimprovementistomakeuseofotherthanTE0modes.Asdiscussed,theprecisionwithwhichdifferentlayerscanberesolvedrequirestheuseofmodeprofilesthathavedifferentfieldstrengthsinthedifferentlayers.Hereweuseddifferentwavelengthstoachievethis.Alternativelyone can use dual polarizations [27], however it will not be possible to distinguish


between specific binding, non‐specific binding, bulk changes simultaneously usingonly two polarizations. On the other hand, higher order modes can be used. Thedifferencebetweenmodeprofilesofe.g.TE0andTE1atthesamewavelengthislargerthanthedifferencebetweenmodeprofilesofe.g.aTE0modeatdifferentwavelengths.Measuring the phase changes for different ordermodes at the samewavelength ispossible[28]butislesstrivialthandetectingthesamemodeatdifferentwavelengths.Alternatively it ispossibletouseaTE0modeatonewavelengthandaTE1modeatanotherwavelength.Thecombinationofmodesand/orwavelengthsthatgivethebestresults requires a systematic experimental investigation and is beyond the scopeofthecurrentpaper.

Currentlymethodstoincreasespecificityofthesensorareaimedtoreducenon‐specificbindingbychemical treatmentof thesensorsurface (e.g.blockingofvacantpositions by bovine serum albumin). The method described here can be used inaddition to thisblockingapproach to further improve the specificity.Approaches toeliminatebulkRIchangesalsoexist.Usuallyafter interactionoftheanalytewiththesensorsurface,thesamplefluidisreplacedbyacleanbufferandthebulkcontributionis obtained by comparing the RI before and after fluid replacement. However thisapproach has some disadvantages. First, no significant changes in binding of theanalytetothesensorsurfaceshouldoccurduringthetimeoffluidreplacement,whichusually means long measuring times. Second, this method is prone to errors as achange fromsample tobuffersolutionsetsanewequilibriumof the freeanalytevs.surface‐bound analyte. Also quite common in interferometric sensors is to use thereferencechanneltoeliminatecommonRIchanges.Inthiscaseareferencechannelismodified identically to the sensing channel without the analyte specific antibody.Applying thesample tobothchannelssimultaneouslystronglyreduces thebulkandnon‐specificcontributions,howeverinpracticethisapproachisnotalwayssufficientespeciallyincomplexmatriceslikeblood.Alsoherethesize‐selectiveapproachcanbeused either alone or in addition to a reference channel to further increase thespecificity.

2.4.Conclusions

In this chapter we have described a new approach, based on the use of multiplewavelengthsincombinationwithanintegratedopticalYounginterferometersensor,whichallowsdetectinganalytesbasedonsize.Theuseofmultiplewavelengthsallowsdiscriminating between RI changes from different locations. To simultaneouslydistinguish between specific binding, non‐specific binding and bulk RI changes, athree layer systemwith threewavelengths isused.Therequiredprecisionof theRIdeterminesthespatialresolutionofthisnewmethod.Assumingaphaseprecisionof10‐4fringes,andananalyteofsize80nmdiameter,thisnewmethodhasaminimum

38 Chapter2

detectablemasscoverageof4×102fg/mm2.Withabetterphaseprecisionitiseven

conceivable to reach a higher precision of n or to introduce an extra imaginarylayer, resulting in a possibility to distinguish from another type of particle with adifferentsize.Furthermore,applyingthisnewmethodtothecurrentsensorgainsinspecificitywhilethedetectionlimitisstillcomparabletothedetectionlimitofotherexistingmethods. Therefore,we believe that thismethod can strongly improve theperformanceofIOinterferometric.

Acknowledgements


Appendix2.AChromaticdispersion

Chromaticdispersion is included in thedeterminationof the sensitivity coefficients.The sensitivity coefficients are determined by themode profile of the light source,

whichdependsonthewavelengthofthelightandthewaveguidestructure( cored , sn ,

coren , cn ). The refractive indices of the chosen materials as a function of the

wavelength are given by the Sellmeier equations, which are listed below in Table2.A.1. Chromatic dispersion in the formation of a layer of particles (e.g. viruses,proteins)onthesurfaceisassumedtobenegligible.Table2.A.1:SellmeierequationsforthedifferentmaterialsusedinthewaveguidestructureMaterial Sellmeierequation Reference

Al2O3 1 22 2 2

2 2 2 2 2 2( ) 1A C E

nB D F

,in[μm]

A=1.4313493, B=0.0726631, C=0.65054713, D=0.1193242, E=5.3414021,

F=18.028251

[22]

Si3N4 1 22

2 2( ) 1A

nB

,in[μm]

A=2.8960,B=0.14010

[23]

TiO2 1 2

2( ) 5.913A

nB

,in[μm]

A=0.2441,B=0.0803

[22]


Material Sellmeierequation Reference

SiO2 1 22

2 2( ) 1A

nB

,in[μm]

A=1.1008,B=0.094025

[23]

Water(20°C)

1 22 2 2 2

2 2 2 2( ) 1A C E G

nB D F H

,in[μm]

A=5.684027565·10‐1, B=5.101829712·10‐3, C=1.726177391·10‐1,

D=1.821153936·10‐2, E=2.086189578·10‐2, F=2.620722293·10‐2,

G=1.130748688·10‐1,A=1.069792721·101

[24]

Appendix2.BDerivationsensitivitycoefficient

In this section we derive the sensitivity coefficient Si,j . Based on a three layerwaveguide with a substrate (ns), a core (ncore), and a cladding (nc), the sensitivitycoefficientisgivenbytheNeffdependenceonaRIchangeintheregionprobedbytheevanescentfield(cladding)andcanbedeterminedforTEpolarizationby[29]:

2 2

0

2 20 0

( )

( ) ( )j

eff core eff c jc

c eff core c j s jcore c

N n N Ynn N d Y Yn n

, (2.B.1)

wheredcoreisthecorethickness. 0cY isthepenetrationdepthoftheelectricfieldinto

thecoveringregionontopofthewaveguide,and 0sY isthepenetrationdepthofthe

electricfieldintothesubstrate,whicharegivenby:

1/22 20 , ,( )

2c s j eff c sY N n

, (2.B.2)

whereλthevacuumwavelengthoftheguidedlight.Thesensitivityofthe thi layerintheevanescentregioncannowbecalculatedasafractionofthecompleteevanescentmodepowerpresentisthislayer[30]:

0

1

0

2( )

, 2( )

0

d

d

i

c j

i

jc j

zzY

z effi j z

cY

e zN

Sn

e z

, (2.B.3)

40 Chapter2

where 0z =0,1

i

i nn

z d

,where nd isthethicknessoflayern.Thisfinallyresultsinthis

expressionforthesensitivitycoefficientforthe thi layerandthe thj wavelength:

1

0 0

2 2 2 2( ) ( ) 0

, 2 20 0

( )

( ) ( )

i i

c j c j

z z

Y Y core eff c jci j

eff core c j s jcore c

n N YnS e e

N d Y Yn n

.(2.B.4)

Appendix2.CDerivationrelativeprecision

HerewederiveanexpressionfortherelativeprecisionΦ ,whereΦ isdefinedasthe

standarddeviationintheRIchangein

normalizedbythestandarddeviationofthe

measuredphasechange.Fromequation(2.4)ofthemaintextwehave:

1

sn M

. (2.C.1)

Thevarianceof n isgivenby:

1 1

2 2s sn M M

, (2.C.2)

where istheHadamardproduct,whichisanelement‐wisemultiplication:

1 1 21

,s s s ijij

M M M

. (2.C.3)

So,the thi elementofvector n canbedescribedby:

1 21 2

1 1 22 1 2,

1i j

N

s sn s iji ji

n M M M

. (2.C.4)

Byassuminganequalphasenoiseforeverylaser(j ),itispossibletodefine

the thi element oftherelativeprecisionas:

1 2 1 2

2 21 2 1, ,

1 1

1i

N Nn

i s ij s ijj j

M M

. (2.C.5).


Appendix2.DDerivationsurfacemasscoverage

WederivehereasetofRIchangesintoaspecificsurfacemasscoverage sC .Ontopof

thecoreofthewaveguidethreeimaginarylayersaredefined.TheRIchangeinlayerone 1n ,iscausedbynon‐specificbindingofparticles(e.g.proteins),specificanalyte

binding(e.g.viruses)andbulkchanges.SpecificbindingandbulkchangesresultinanRI change in layer two and the RI change in layer three is only caused byconcentration changes of the bulk. In addition, we convert the RI change of the

different layers into a non‐specificmass coverage, nC , a specificmass coverage, sC ,

andaconcentrationchangeofthebulk bC .Therefore,therelationbetween nC , sC , bC

and n isgivenby:

1

2

3

, 0

0 0

n

C s C

b

n C

n M C with M

n C

, (2.D.1)

where , , and , convert an RI change into respectively a non‐specific mass

coverage,aspecificmasscoverageandabulkconcentrationchange.Weassumesmallconcentration changes, sowe can neglect changes in , , and , because of the

growthofanalytestothesurface,whichresultsinlessspaceonthesurfacewhichcanbecoveredbynewanalytes.Thenon‐specificmasscoverage,specificmasscoverageandbulkconcentrationchangecanbecalculatedby:

11 1

2

3

1 1 0

, 0 1 1

0 0 1

n

s C C

b

C n

C M n with M

C n

. (2.D.2)

Fromthisthemasscoverage sC isgivenby:

2 3

1( )sC n n

, (2.D.3)

where 2n is composed of two contributions: 2 specific bulkn n n , whereas

3 bulkn n ,whichresultsinanequationformasscoverage:

1

s analyteC n

. (2.D.4)

42 Chapter2

Next,wedetermine .Wedefineanexpressionfor sC as:

analyte analytes

N mC

w l

, (2.D.5)

where analyteN isthenumberofanalytesboundthesurface, analytem themassofasingle

analyte, w the sensitivity window width (= 4 μm for current chip), and l thesensitivity window length (= 4 mm for current chip). The maximum number of

analytes maxN that can be bound to the surface is determined as the ratio of the

surface area of the sensor and the surface area of a single analyte and can be

approximatedas 2maxN w l d ,wheredisthediameteroftheanalyte.Ifwedefinea

layer of thickness equal to the diameter of the analyte, than the corresponding

maximum RI change maxn in that viral layer equals max analyte solutionn n n . If in an

experimenta n is introducedbyspecificanalytebinding(denotedas analyten ), the

numberofanalytes analyteN boundtothesurfaceofthechipiscalculatedas:

max 2

max

analyte analyteanalyte

analyte solution

n n w lN N

n n n d

. (2.D.6)

CombiningEq.(D.4),Eq.(D.5),andEq.(D.6),thisresultsinanexpressionfor :

2analyte solution analyte virusn n d m . (2.D.7)

BasedonEq.(D.4)(so 2 3( , )sC f n n )theminimaldetectablespecificmasscoverage

isgivenby:

1 22 2

1 22 2

2 3 2 32 3

1

v

f fC n n n n

n n. (2.D.8)

Ifweassumethattheerrorin 2n , 2n ,isgivenbytheprecisionin 2n (relative

precision multiplied by the precision in the phase change),2n

, and that

2 analyte bulkn n n ,thantheminimaldetectablespecificmasscoverageofananalyte

canbeapproximatedby:

2

analytev nC . (2.D.9)


References

1. A. Brandenburg, R. Krauter, C. Kunzel, M. Stefan, and H. Schulte, “Interferometricsensorfordetectionofsurface‐boundbioreactions,”Appl.Opt.39,6396‐6405(2000).

2. K. Schmitt, B. Schirmer, C. Hoffmann, A. Brandenburg, and P. Meyrueis,“Interferometricbiosensorbasedonplanaropticalwaveguidesensorchipsforlabel‐free detection of surface bound bioreactions,” Biosens. Bioelectron. 22, 2591‐2597(2007).

3. C.Stamm,R.Dangel,andW.Lukosz,“Biosensingwiththeintegrated‐opticaldifferenceinterferometer:dual‐wavelengthoperation,”Opt.Commun.153,347‐359(1998).

4. F. Prieto, L. M. Lechuga, A. Calle, A. Llobera, and C. Dominguez, “Optimized siliconantiresonant reflecting optical waveguides for sensing applications,” J. LightwaveTechnol.19,75‐83(2001).

5. R. G. Heideman, R. P. H. Kooyman, and J. Greve, “Performance of a highly sensitiveoptical waveguideMach‐Zehnder interferometer immunosensor,” Sens. Actuators, B10,209‐217(1993).

6. A. Ymeti, J. S. Kanger, R. Wijn, P. V. Lambeck, and J. Greve, “Development of amultichannel integrated interferometer immunosensor,” Sens. Actuators, B 83, 1‐7(2002).

7. P.Kozma,A.Hámori,S.Kurunczi,K.Cottier,andR.Horvath,“Gratingcoupledopticalwaveguideinterferometerfor label‐freebiosensing,”Sens.Actuators,B155,446‐450(2011).

8. K. Cottier, M. Wiki, G. Voirin, H. Gao, and R. E. Kunz, “Label‐free highly sensitivedetection of (small) molecules by wavelength interrogation of integrated opticalchips,”Sens.Actuators,B91,241‐251(2003).

9. M. S. Luchansky,A. L.Washburn, T.A.Martin,M. Iqbal, L. C. Gunn, andR. C. Bailey,“Characterizationoftheevanescentfieldprofileandboundmasssensitivityofalabel‐freesiliconphotonicmicroringresonatorbiosensingplatform,”Biosens.Bioelectron.26,1283‐1291(2010).

10. G.D. Kim, G. S. Son,H. S. Lee, K.D. Kim, and S. S. Lee, “Integrated photonic glucosebiosensorusingaverticallycoupledmicroringresonatorinpolymers,”Opt.Commun.281,4644‐4647(2008).

11. M.Iqbal,M.A.Gleeson,B.Spaugh,F.Tybor,W.G.Gunn,M.Hochberg,T.Baehr‐Jones,R.C. Bailey, and L. C. Gunn, “Label‐Free Biosensor Arrays Based on Silicon RingResonatorsandHigh‐SpeedOpticalScanningInstrumentation,”IEEEJ.Sel.Top.Quant.Electron.16,654‐661(2010).

12. N.Skivesen,A.Tetu,M.Kristensen,J.Kjems,L.H.Frandsen,andP.I.Borel,“Photonic‐crystalwaveguidebiosensor,”Opt.Express15,3169‐3176(2007).

13. S. C. Buswell, V. A. Wright, J. M. Buriak, V. Van, and S. Evoy, “Specific detection ofproteinsusingphotoniccrystalwaveguides,”Opt.Express16,15949‐15957(2008).

14. A.Ymeti,J.Greve,P.V.Lambeck,T.Wink,S.W.F.M.vanHovell,T.A.M.Beumer,R.R.Wijn, R. G. Heideman, V. Subramaniam, and J. S. Kanger, “Fast, ultrasensitive virusdetectionusingayounginterferometersensor,”NanoLetters7,394‐397(2007).

15. C. Worth, B. B. Goldberg, M. Ruane, and M. S. Unlu, “Surface desensitization ofpolarimetricwaveguideinterferometers,”IEEEJ.Sel.Top.Quant.Electron.7,874‐877(2001).

16. A. Ymeti, “Development of a multichannel integrated Young interferometerimmunosensor,”Ph.D.Thesis,UniversityofTwente(2004).

17. J. Homola, “Surface plasmon resonance sensors for detection of chemical andbiologicalspecies,”ChemicalReviews108,462‐493(2008).

44 Chapter2

18. A.Ymeti,J.S.Kanger,J.Greve,P.V.Lambeck,R.Wijn,andR.G.Heideman,“Realizationof a multichannel integrated Young interferometer chemical sensor,” Appl. Opt. 42,5649‐5660(2003).

19. E.Hecht,inOptics(Addison‐Wesley,NewYork,1998),pp.385‐388.20. W.M. Balch, J. Vaughn, J. Navotny,D. T.Drapeau, R. Vaillancourt, J. Lapierre, andA.

Ashe,“Lightscatteringbyviralsuspensions,”Limnol.Oceanogr.45,492‐498(2000).21. M. Green, M. Piña, R. Kimes, P. C. Wensink, L. A. MacHattie, and C. A. Thomas,

“AdenovirusDNA.I.Molecularweightandconformation,”Proc.Natl.Acad.Sci.U.S.A.57,1302‐&(1967).

22. W.J.Tropf,Thomas,M.E.,Harris,T.J.,“Propertiesofcrystalsandglasses,”inHandbookofOptics:Fundamentals,techniques,anddesign(McGraw‐Hill,1995),pp.33.61‐33.66.

23. T. Bååk, “Silicon oxynitride; a material for GRIN optics,” Appl. Opt. 21, 1069‐1072(1982).

24. M.DaimonandA.Masumura,“Measurementoftherefractiveindexofdistilledwaterfrom the near‐infrared region to the ultraviolet region,” Appl. Opt. 46, 3811‐3820(2007).

25. P.Kozma,F.Kehl,E.Ehrentreich‐Förster,C.Stamm,andF.F.Bier,“Integratedplanaroptical waveguide interferometer biosensors: A comparative review,” Biosens.Bioelectron.58,287‐307(2014).

26. J.Dostálek, J.Čtyroký, J.Homola,E.Brynda,M.Skalský,P.Nekvindová, J.Špirková, J.Škvor, and J. Schröfel, “Surface plasmon resonance biosensor based on integratedopticalwaveguide,”Sens.Actuators,B76,8‐12(2001).

27. G. H. Cross, A. A. Reeves, S. Brand, J. F. Popplewell, L. L. Peel,M. J. Swann, andN. J.Freeman,“Anewquantitativeopticalbiosensorforproteincharacterisation,”Biosens.Bioelectron.19,383‐390(2003).

28. K.E.Zinoviev,A.B.Gonzalez‐Guerrero,C.Dominguez,andL.M.Lechuga,“IntegratedBimodalWaveguide InterferometricBiosensor forLabel‐FreeAnalysis,” J. LightwaveTechnol.29,1926‐1930(2011).

29. P.M.NellenandW.Lukosz,“Integratedopticalinputgratingcouplersasdirectaffinitysensors,”Biosens.Bioelectron.8,129‐147(1993).

30. E. F. Schipper, “Waveguide immunosensing of small molecules,” Ph.D. Thesis,UniversityofTwente(1997).

Chapter3

Design,realizationandcharacterizationofasize‐selectiveYounginterferometersensorAbstractThis chapter presents the design, realization, and characterization of a Younginterferometer sensor setup capable tomeasure simultaneously effective refractiveindex changes at multiple wavelengths to achieve size‐selectivity. First, therequirements and an overview of the setup are presented. Next, we describe therealization and characterizationof the sixmainmodulesof the setup: light sources,incoupling, sensing platform, imaging, detection and data processing. Finally, wecharacterise thephasenoise anddrift of the setup (unexpectedphase changeson alongertimescale).Itwasfoundthatthemeasuredphasenoiseonshorttimescalesisdetermined by photon shot noise. A CCD camera with a high dynamic range wasimplemented in the setup such that themeasured phase noise is smaller than 10‐4fringes@ 1 Hz.Phasedriftissmallerthan5x10‐3fringesper1000sforthesetupwithend‐firecoupling.The incouplingwasmademoretimeefficientbyusinga fiberandbutt‐endcoupling.However,thedriftincreasedtovaluessmallerthan1.5x10‐2fringesper 1000 s, partly caused by positional drift of the fiber. Drift in distance betweenfiberandchipresulted inphaseoscillationswhichcouldbesolvedusingamatchingindexgel.Itwasalsofoundthatairflowcanstronglyinfluencethephasesignal,sothesetupwascovered.Moreover,changesintheambienttemperaturehaveaneffectonthephasestabilityasitcanleadtothermalexpansionofcomponentsofthesetupandthechipitself.Thiscanleadtochangesinthepositionofthechipwithrespecttothecamerawhichresultsinphasechanges.Nexttodriftalsoartefactsareobservedinthephase signal. The origin of these artefacts is investigated in this chapter. Possiblereasonsoftheartefactsareaberrationsofgratingandlensesintheimagingpartofthesetupandboundaryeffectsoftheshiftinginterferencepattern.Theartefactsshowupas oscillations in the phase change with an amplitude in the order of 10‐2 fringes.Inducedsignalsshouldbesignificantlyhigherthandriftandartefactstominimizetheinfluenceofdriftandartefactsonthedetermined n ’s.

46 Chapter3

3.1Introduction

Tomeasure size‐selectivelywith the YI usingmultiplewavelengths there are somerequirementsforthesetup.First,todiscriminatebetweenthreedifferentRIchangesat least three differentmonochromatic coherent light sourceswith each a differentwavelength are required. Second, these light sources should be coupled into anexistingintegratedopticalYI.Third,aftercouplingoutoftheYI,thelightsourceswillformthree interferencepatternswhichshouldbedetected independently fromeachother. Fourth, the three interference patterns should be read out and fast Fouriertransform(FFT)shouldbeappliedtoeachofthemtodeterminephasechanges( (

~ effN ),seeformula2.1)foreachwavelengthandeachchannelcombination.Fifth,

Chapter2showedthatwithaddingsize‐selectivitytotheYI,thesensitivityoftheYIisnegativelyaffected.Anynoisein effN willbeenhancedbydeterminingmultiple n

’sfrom effN ’sdeterminedfrommeasured ’satmultiplewavelengths.However,

Chapter 2 also showed that if ismeasuredwith a precision of 10‐4 fringes the

detection limit of the YI in combination with size‐selective detection is stillcomparable to the detection limits of other existingmethods. Therefore, the phasenoise shouldpreferably be as lowas possible and in the orderof10‐4 fringes to becomparabletootherexistingmethodsintermsofsensitivity.

A setup was designed, realized and characterized such that it meets all therequirementsnamedabove.Forconvenienceofcomparison,thesetupisdivideduptosixmainpartswhicharepresentedinafunctionblockdiagramasshowninFig.3.1.Eachofthesections3.2–3.7belongstoasingleblockinthefunctiondiagram.Next,section 3.8 presents a final overview of the self‐built setup. Finally, section 3.9presents experiments whichwere done to characterize phase noise (on short timescaleof seconds)andphasedrift (on longer timescaleofhundreds to thousandsofseconds)oftheself‐builtsetup.

Fig.3.1:Schematicoverviewofthesetuppresentedasafunctionblockdiagramillustratinginwhichsectionwhichpartofthesetupisdescribedandthatthefinaloverviewispresentedinsection3.8.

Design,realizationandcharacterizationofasize‐selectiveYounginterferometersensor 47

3.2Sensingplatform

3.2.1Four‐channelintegratedopticalYounginterferometerFirst,wedescribe thesensingplatformof thesetup,whichwasdeveloped inearlierwork [1], for the size‐selective measurements using multiple wavelengths. Thesensing platform is a four‐channel integrated optical Young interferometer. Light isfirstcoupledintoasinglechannelwithacorewidthof10μmatthebeginningofthechipfollowedbya100μmlongtapertoafinalwidthof4μm.Next,thesinglechannelsplitsupintofourdifferentchannels.Thewaveguideconsistsofthreelayers,aPECVDSiO2 layer, on top of a 70 nm thick LPCVD Si3N4 core, on top of a Thermal SiO2substrate. A schematic top view and lateral view of the end part of the YIinterferometerareshowninFig.3.2aandbrespectively.TheYIisaridgewaveguidewhichmeansthat lateralconfinementofthechannels isrealizedbythe0.7nmhighand 4 μm wide ridges shown in Fig. 3.2b. Therefore, the modes propagate at the

ridgesresultinginfourchannelsdistancedat60μm,80μmand100μm.Thedistancesof thechannelsarechosensuchthat thereare6differentcombinationsofdistancesbetween thechannels, resulting in6differentspatial frequencieswhichcanbereadout individually from the interference pattern after applying an FFT on it. At thesensingwindowsofthefourdifferentchannels,thetoplayerofPECVDSiO2isetchedawaysuchthatthesamplecanreachthecoreofthewaveguide.Across‐sectionofthewaveguideatthesensingwindowsispresentedinFig.3.3.

TheYI chipshave a singlemode (inheight)wavelength rangeof approximately170 ‐ 800 nm. However, the chips are optimized for TE modes of 632.8 nm light.Therefore,more losses canbe expected for especially shorterwavelengths than theoptimal 632.8 nm. Incoupling will be more critical for shorter wavelengths as themode size is smaller for shorter wavelengths.Moreover, propagation losses due toscattering(SiO2andSi3N4 layersareamorphousandcanhavesmalldefects)willbehigher for shorter wavelengths (~ 4 ) and at the y‐splitters there might be someextra losses for smallerwavelengthwhichhavesmallerpropagationmodes.At 632.8 nm only TE00 or TM00 modes can propagate through the waveguides. Wecalculatedusinganeffective indexmethod[2] thatthisalsoholds for thecurrentYIwaveguide at wavelengths longer than 540 nm. However, for shorter wavelengthsthan 540 nm the TE01 and TM01 do also fit in thewaveguide. Thismeans that inheightthepropagatingmodesaresinglemode,butinwidth,thezerothandfirstordermodefitintothewaveguide.Usingthesameeffectiveindexmethodwecalculatedthatthesensitivitiesofzerothandfirstorderarealmostequal.Theeffectiveindexmethodhas limitations, but for a weak lateral confinement, which is the case with thiswaveguide,theprocedureisusuallywelldefined[2].Consequently, themultimodesshould not give any problem and the sensitivities of the TE00 and TE01mode areassumed equal.However, for the incoupling themultiplemodesdomatter, because

48 Chapter3

thepositionof the zeroordermode (from the fiberor focussedwith lens)which iscoupled into the waveguide does determine if mainly TE00 or TE01 modes arecoupled into thewaveguide.Thiswill result innon‐equaldistributionof the light inthechannelsoccurringatthey‐splitters.Detailsabouttheincouplingwillbediscussedinsection3.4.

Fig.3.2:a)Topviewillustratingthesensingwindows(blackrectangles)andpositionof fluidcuvette(blackdottedrectangles)andb)lateralviewoftheendpartwaveguidestructure,wherethe0.7nmand4 μm ridgesdefinethelateralconfinementofthelightandcanthereforebeseenaschannels.Imagesarenottoscale.

Fig.3.3:Cross‐sectionoftheYIwaveguidestructureguidingaTE0modeof660nm.Inheightthe image is to scale,whichmeans that the evanescent fieldpenetrationdepth is a coupleofhundrednanometres.


3.2.2SampletransportationTobringthesampletotheevanescentfieldofoneofthefoursensingwindowsoftheYIwaveguide,a4channelfluidcuvettewitheachasingleinletandoutletispositionedontopofthewaveguideasillustratedwiththeblackdottedrectanglesinFig.3.2.Thefluidcuvetteconsistsoffourflowchambers,eachofavolumeof1.2μl(2mmwide,6mmlongand100μmhigh).ThesampleisbroughttothesensingwindowsoftheYIusing a 4 tube Reglo Digital MS‐CA4/12‐100 peristaltic pump with 0.57 mm innerdiameterTygontubing.Insidethecuvettethesampleistransportedbylaminarflow.As a consequence, the velocity at the surface is zerowhichmeans that the analyteswillbetransportedtothesensorsurfacebydiffusionperpendiculartotheflow.Theflowspeedduringthemeasurements issetat100μl/min.Beforethesamplesreachthesensingwindowsitisdegassedbya4channelBiotechDegasi®Classictopreventunwantedairbubbleswithcanresult inahigh n onaveryshorttimescalewhichcansignificantlydisturborevenruintheexperiment.

3.3Lightsources

3.3.1RequirementsAs stated before, the light sources that are required for the YI should bemonochromaticandcoherent.Thelightsourceshouldbeasingletransversemodetoeffectively couple the light into the YIwaveguide. For a stable incoupling, the lightsource should also have a stable beam pointing. Furthermore, a single longitudinalmodewithahighlevelofstabilityisrequiredtopreventmode‐hoppingwhichresultsin unstable intensity andwavelength and therefore anunstable phase signal. To beable to measure an interference pattern and stable phase signals the light sourceshouldhaveasufficientlylongcoherencelength.Besides,calculationsfromChapter2showthatbestresultsareobtainedusinganaslargeaspossiblespreadbetweenthewavelengths.Ontheotherhand,thespreadislimitedbytheYIwaveguidestructure,which should only be guiding zero order modes of the light source of specifiedwavelength.

3.3.2RealizationTheCoboltdiode‐pumpedsolidstate(DPSS)lasershaveanexcellentperformance,intermsofpowerstability,wavelengthstability,beampointingandcoherencelength[3,4] and therefore the (457.0±0.3) nmCoboltTwistTM and the (561.2±0.3) nmCoboltJiveTMfromthe(660±0.3)nmCobolt04‐01SeriesandtheCoboltFlamencoTMfromthe05‐01 Cobolt serieswere implemented in the experimental setup. Thewavelengthswhere chosen such that they arewell spread and that single ordermodes of thesewavelengthsdostillfitintothewaveguide.Lightsourcesof50mWwerechosentobe

50 Chapter3

abletocompensateforpossiblyhighcouplinglossesintothewaveguide,especiallyforshorter wavelengths than the 632.8 nm for which the waveguide was optimized.ThorlabsNDC‐100C‐4MmountedvariableNDfiltersareusedtoregulatethepowerofthe transmitted light of the light sources as the DPSS laser perform optimally atmaximumoutputpower.Acoupleofminutesarerequiredforthelasertostabiliseitstemperature and to achieve power, wavelength and beam pointing stability whichmakesmodulatingof the laser impracticable.By implementing these lasers into thesetupall requirementsof the light sourceswere fulfilled tomeasuresize‐selectivelyusingmultiplewavelengths.

3.1.3CharacterizationThepowerofthelaserswasverifiedusingaThorlabsPM121Dpowermeterandthepolarizationof the laserwasverifiedusingaThorlabsLPVISE100‐A linearpolarizerwithN‐BK7 ProtectiveWindows (400‐700 nm). Themeasured data is presented inTable3.1.Thesepowersweresufficienttoovercomecouplinglossesandstillbeableto detect the interference patterns coupled out of the chip. Moreover, withpolarizationratios(TM:TE)largerthan100:0.009itispossibletocouplewelldefinedpolarized modes into the waveguide, which is required as modes of differentpolarizations respond differently on RI changes in the evanescent field (seeChapter6).

Phaseoscillationswereobservedat the660nm interferencepattern, indicatingthattheoscillationswerecausedbythe660nmlaser.AfterweekswefoundtogetherwiththeCoboltcompanythatthetemperaturecontrollersof the660nmlaserwerecausing the oscillations. Overcompensating the temperature by the laser causestemperature oscillations possibly resulting in oscillations in wavelength, beampointing or power of the laser. By loweringmultipliers of the PID controller of thelasertheseoscillationswerenotseenagainsothisproblemsolved.Table3.1:Measuredspecificationsofthelightsources CoboltTwistTM CoboltJiveTM CoboltFlamencoTM

Power(mW) 49.5 49.5 57.7PTE(mW) 0.360 0.115 0.151PTM(mW) 38.0 38.8 46.5Polarizationratio(TM:TE)

100:0.009 100:0.003 100:0.003

3.4Incoupling

3.4.1RequirementsThelightcomingfromthethreelasersshouldbesimultaneouslycoupledintoaridgewaveguidewitha(70±1)nmthickcoreandaridgeof0.7nmhighand10μmwide


(which is tapered to a width of 4 μm). The linear polarization of the light sourceshouldbemaintainedand itsorientationwith respect to thewaveguidedetermineswhether TM (perpendicular to core layer, vertical) or TE (parallel to core layer,horizontal)modesarecoupledintothewaveguidestructure.

3.4.2RealizationEnd‐firecoupling,butt‐endcoupling,prismcoupling,gratingcouplinganddirectionalcouplingaredifferentways tocouplethe lightsources into thewaveguide[5].First,theincouplingwasrealizedusingopticsinfreespacetooverlapthethreelasersandusinganachromatic lens to focus the light into the core structureof thewaveguide(end‐fire coupling) to be flexible in adding or removing components in the setup.Lateron in theproject, the lightwascoupled into thewaveguideusingsinglemodefiberandbutt‐endcouplingtogotowardsafasterandeasierincouplingofthethreewavelengthsandamoreapplicationdrivensolution.Prismcoupling,gratingcouplingand directional coupling were not used because of simultaneously incoupling ofmultiplewavelengthsintothewaveguideandthegivenwaveguidestructureoftheYI.Theend‐firecouplingandthebutt‐endcouplingaredescribed insection3.4.2.1and3.4.2.2respectively.

3.4.2.1End‐firecouplingTo use end‐fire coupling the wavelengths should first be overlapped. A schematicoverviewofhowthiswasdoneisshowninFig.3.4.Thelightsourcesasdescribedinsection3.3arefirstdirectedtoThorlabsNDC‐100C‐4MmountedvariableNDfilterstoregulate the transmitted power per laser. The transmittance can be regulated from1%to100%,becauseofavaryingopticaldensityof0to2.0.Forsafetyreasons,thereflected light iscapturedbyBT500beamdumps.The transmitted light firstpassesThorlabsMountedZero‐OrderHalf‐WavePlates(Thorlabs,WPH05M‐488,WPH05M‐546 andWPH05M‐670) to change the linear polarization of the light sources to beable to couple either TE or TM modes into the waveguide. Next, a set of visiblebroadband dielectric mirrors (Thorlabs, BB1‐E02) and dichroic mirrors (Semrock,Di01‐R561‐25x36(DM1)andDi01‐R442‐25x36(DM2))areusedtooverlapthethreelightsources.DM1hasanaveragetransmissionof94%forwavelengthhigherof582.4nmtill1200nmandanaveragereflectionof94%ofwavelengthsinbetween554nmand 568 nm. Therefore, itwill reflectmost of the light of the 561 light source andtransmit most of the light of the 660 nm light source. The DM2 has an averagetransmission of 93% for wavelengths in between 469.3 nm and 900 nm and anaverage reflection of 94% of wavelengths in between 439 nm and 457.9 nm.Therefore,itwilltransmitmostofthelightcomingfromthe561nmand660nmlaserandreflectmostofthelightcomingfromthe457nmlaserwhichmakesitpossibletooverlapthelaserbeams.Next,thecollinearbeamsarepointedviaamirrorto

52 Chapter3

Fig.3.4:Schematicoverviewoftheend‐firecouplingtocouplethelightintotheYIwaveguide,whereNDstandsforneutraldensityfilter,Mstandsfordielectricmirror,HWPstandsforhalf‐waveplate,DMstandsfordichroicmirror,GMforadielectricmirrorplacedinGimbalmirrormountandLstandsforachromaticlens.the centre of another mirror positioned in a Gimbal mirror mount whose point ofrotation for both axes is located at the centre of the mounted optic’s surface.Therefore, changing the angleof thebeamby tilting thismirror, thepositionof thebeam at the mirror is not changed, which is required to position the beam at thecentreoflensL3.ThemirrorintheGimbalmirrormountispositionedatthefirstfocalplaneofa4flenssystemconsistingoftwoachromaticdoublets(AC127‐019‐A‐ML,f1=

19 mm and AC254‐100‐A‐ML, f2 = 100mm). At the final focal plane of the 4f lenssystema thirdachromaticdoublet (AC254‐040‐A‐ML, f3=40mm) isplaced to focusthelightintothewaveguidestructure.Raytracingshowsthatthexfandyfpositioncanbe adjustedby changing the angleof thebeamwith respect to thehorizontalplane

0 ,x y by(seeAppendix3.A):

3 10 ,

2

,f f x y

f fx y

f

(3.1)

where fn is the focusof thenth lens.Thepositionxfcanbesetbyadjustingtheangle

0x and the position of yf can be set by adjusting 0 y . The lenses L1 and L2 also

functionasabeamexpanderwhichdeterminesthefocalbeamwidthwf,givenby:

1 3

0 2

4f

f fw

w f

(3.2)

where isthewavelengthofthelight,w0thebeamwidthatthegimbalmirror.Thefocal beamwidthwf is calculated to be 7.61 μm, assuming = 550 nm, the focallengthsofthelensesnamedbeforeandw0=700μm(1/e2)whichisthebeamwidthof


the laser. This does not fit perfectly to the singlemode of the waveguide which iseasilyafactoroftentimessmallerinheight,whichcanleadtocouplinglossesaswasstatedbefore. Inwidth, themodeprofile iswiderbecauseof the10μmwide ridge,which leads to a higher coupling efficiency itmeans also that in the beginning firstordermodescanbecoupledintothewaveguide.Couplingofthesefirstordermodesinwidthcanbeminimizedbyfocussingthelight inthecentreofridge.Usingthe50mW lasers, the coupling losseswereovercomeanda sufficient amountof lightwascoupledintoandfinallyoutofthewaveguidetodetectthreeinterferencepatterns.

3.4.2.3FiberbuttcouplingThe incoupling method was changed to butt‐end coupling to go towards a moreapplicationdrivensolutionandamoretime‐efficientincoupling.ThesetupofFig.3.4was maintained up to and including mirror M5 after which an apochromatic fibercollimator (60FC‐4‐RGBV11‐47, Schäfter + Kirchhoff GmbH, Hamburg, Germany)placedonaxyz‐stage(MAX361D/MFiberLaunch,Thorlabs)isusedtocouplethelightintoaThorlabsPM460‐HPsinglemodepolarizationmaintaining(SMPM)opticalfiber.Instead, it is also possible to use fiber pigtailed lasers in combination with anIntegratedLaserBeamCombiner[6]tocoupleallthelightsourcesinonesinglefiberwhich can be positioned with respect to the waveguide to couple light into thewaveguideviabutt‐endcoupling.Aswealreadybought thenon‐pigtailed lasers thiswasnotanoption,butforafinalproductwhichcouldbebroughtontothemarketthisoptionshouldbetakenintoaccount.

Forfiberbutt‐endcouplingthehalf‐waveplatesaretakenoutofthesetupbecausetheyarenotrequiredasthepolarizationof the lightcanbechangedbyrotatingthefiber. The end of the fiber is positioned on top of an ULTRAlign™ Precision XYZPositioningStagewithDS‐4FHighPrecisionAdjusters(8.0mmCoarseTravel,0.3mmFineTravel)topositionthefiberwithrespecttothewaveguideinordertoefficientlycouplethelightintothewaveguide.Thesinglemodefiberhasacorediameterof3.0μmandamodefielddiameterof3.3±0.5μm(1/e2fitnearfield,[7])at515nmandissingle mode above 450 nm. To reach the waveguide which is embedded in a chipholderandtoprotect the fibercore, theendof the fiber isglued intoaglass ferrulewithasizeof1mmandaholeof125μmbyXiOPhotonicsBV.Aftergluingthefiberintheferrule,theferruleandfiberarepolishedtogether.Consequently,theferruledoesnotaffectthemodeprofile.Thecouplingofthesinglemodeofthefiberisespeciallycritical in height as the size of the single mode of the waveguide in height isapproximately0.5μm(1/e2)at515nm.Inwidthitisalsoimportanttoalignthefiberatthecentreofthewaveguidetapersuchthatzeroordermodesarecoupledinmostefficientlytogetthemostequaldistributionofthelightoverthefourchannels.UsingaUSBcameraforthealignmentofthefiberwewereabletocouplethe457nm,561nmandthe660nmlightsimultaneouslyintothefiber.However,changingtheincouplingof the system from end‐fire coupling to fiber butt‐end coupling, oscillations in the

54 Chapter3

phase signal appeared with an amplitude of approximately 3x10‐3 fringes (see Fig.3.5a).Therefore,thephaseprecisionof10‐4fringescouldnotbeachieved.Oscillationswere alsomeasured in themean signal (mean total number of counts of thewholeinterference pattern) of the measured interference patterns. The period of theobserved oscillations in phase and mean number of counts were wavelengthdependent,indicatingthattheyarecausedbyaninterferenceeffect.Usingapiezoxyz‐stagetoactivelychangethepositionofthefiberbyatriangularwavewithrespecttothewaveguidetheoriginoftheoscillationswastracedtoachangeindistanceofthefiber with respect to the chip. Similar oscillations in phase and mean counts aremeasuredwhenchanging thedistance (z‐direction)andnot for changing theheight(y‐direction) and width (x‐direction) as shown Fig 3.5b‐d. The period of theoscillationsshowninFig.3.5darenotconstantovertime,whichmeansthatthedriftisnotconstantovertime.Thiswasnotexpectedbecauseoftheuseofthetriangularwave.However,thiscanbeexplainedbythefactthatthepiezostagewaspositionedon topof another stagewhichwasused todo the coarse alignment andwhich alsoslightly drifts with a non‐constant velocity over time. The oscillations in themeannumberofcountscanbeexplainedbythefactthatthefibertipandwaveguidebehavelike two partly reflective mirrors in a Fabry‐Perot cavity. It was expected that thephasesignalwouldnotshowoscillations,becausetheincomingbeamiscoupledintoone channel which splits finally up into four channels in which beams shouldpropagatewiththesamephase,independentoftheincomingphaseorintensity.Slablightwhichisnotcoupledintothechannelsbutiscoupledoutattheendfaceofthechipmightbeareasonwhyoscillationsarealsoseeninphasesignal.

Oscillationscanbepreventedbyusingaxyz‐stagewithasufficientlysmalldrift.Over 2000 s the fiber was displaced 2 μm and approximately 13 oscillations weredetected for the 660 nm laser. This means that one oscillation corresponds toapproximately154nm.Consequently,foramaximumphasenoiseof10‐4fringesandthe measured amplitude of approximately 3x10‐3 fringes, only a thirtieth of anoscillation is allowed. This means that the stage should not drift more thanapproximately5nm.Ontheotherhand,thestageshouldbeabletodisplacethefiberfor a couple of millimetres. It is hardly possible to find a stage whichmeets theserequirements. Therefore, the problem was solved placing an index matching gel,which matches the refractive index of the glass fiber tip, in between fiber andwaveguideandbringthefiberincontactwiththewaveguide.Therefore,thelightwillnot be back reflected at the interface of fiber andmatching gel. An indexmatchingfluidalsowassufficienttogetridofoscillationsbutdidleakintothemeasurementschambersof thewaveguidestructureandwas thereforenotsuitable.Fig.3.6showsthatusingamatchingindexgeltheoscillationsinthephasesignaldisappeared.


Fig.3.5:a)Phasechangeandmeannumberofcountsofinterferencepatternsmeasuredat =561nmand660nmshowing typical oscillations and thephase changeandmeannumberofcountswhenchangingpositionof the fiber inb)y‐direction (height), c) x‐directionandd) z‐direction(distancetothechip).

Fig.3.6:a)Phasechangeandmeannumberofcountsfora)fibernotincontactwithwaveguide,b)fiberincontactwithwaveguide,c)fiberincontactwithwaveguidewithindexmatchinggelinbetween.

0 1000 2000 3000‐0.5

0.0

0.5

1.0

1.5

2.0

2.5

0 100 200 300 400

0

1

2

3

4

5

0 50 100 150 200 250‐5

‐4

‐3

‐2

‐1

0

1

2

0 500 1000 1500 2000‐0.5

0.0

0.5

1.0

1.5

dc

bPhasechange(fringes)

Time(s)

660nm,d12

561nm,d23

ax10‐2 x10‐5

0.5

1.0

1.5

2.0

2.5660nm561nm

Meancounts(a.u.)

Phasechange(fringes)

Time(s)

660nm,d12

561nm,d23

2

3

4

5

6

7660nm561nm

Meancounts(a.u.)

x10‐2 x10‐5


Time(s)

660nm,d12

561nm,d23

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0x10‐5

Meancounts(a.u.)

660nm561nm

x10‐2


Time(s)

660nm,d12

561nm,d23

2.0

2.5

3.0

3.5

4.0

Meancounts(a.u.)

660nm561nm

x10‐2 x10‐5

0 1000 2000 3000‐6

‐4

‐2

0

2

4

6

0 1000 2000 3000‐8

‐6

‐4

‐2

0

2

4

0 1000 2000 3000‐10

‐8

‐6

‐4

‐2

0

2cb


Time(s)

660nm,d12561nm,d

12

x10‐2 x105a

‐2

0

2

4

6

8

10660nm561nm

Meancounts(a.u.)


Time(s)

660nm,d14561nm,d14

‐2

0

2

4

6

8

10

Meancounts(a.u.)

x10‐2 x105

660nm561nm


Time(s)

660nm,d12561nm,d

12

x10‐2 x105

0

2

4

6

8

10

12

660nm561nm

Meancounts(a.u.)

56 Chapter3

Sousingthematchingindexgelandthebutt‐endcouplingasufficientamountoflightwascoupledintoandoutoftheYIwaveguidetomeasureaphasechangeofthethreeinterference patterns without the oscillations such that this does not restrict therequiredphaseprecisionof10‐4fringes.

3.4.3CharacterizationThelaserswerecoupledintoaPM460‐HPSMPMopticalfiber.Thisfiberisa“Panda”stylePMfiberwhichmeansithasafastandaslowaxiswhichshouldbealignedwiththe polarization of the light such that fiber is polarization maintaining. If it is notalignedproperly,thereisalsoasmallTEcomponentwhichpropagateswithaslightlydifferent velocity and then the length of the fiber determineswhat fraction TE/TMcomesoutofthefiber.Alsobendingofthefiberandtemperaturechangescanchangethe output, because the velocity of the TE and TMmodes are changed differently.Coupling efficiencies achieved using the 60FC‐4‐RGBV11‐47 apochromatic fibercollimatorwere37%forthe457nmlaser,51%forthe561nmlaserand50%forthe660nmlasercouplingall thethree lasers intothe fiber.Highercouplingefficienciescouldbereachedbycouplingasinglelaserintothefiber.UsingtheMAX361D/MFiberLaunch xyz‐stage the incoupling was very stable after settling of the adjustmentscrews. No further adjustment was required after once optimising the incoupling.Laser light with a ratio TE/TM of 100:1 was coupled into the fiber. ThemeasuredTE/TMratiooutofthefiberwasvariedfrom1:100to5:100forallthelasers.TheTMpolarizedlightwasfilteredoutafterthewaveguideusingapolarizer.TheTE/TMratiowas strongly influenced by the bending of the fiber. Therefore, the fiber should bestablypositionedduringthemeasurements.Bystablepositioningthefiberandusingandpolarizerbehindthewaveguide,wewereable tomeasurephasechangesof thethreewavelengthsofeitherTEorTMpropagatingmodes.

3.5Imaging

3.5.1RequirementsTomeasure size‐selectivelywithmultiplewavelengths it is necessary to detect theinterferencepatternsof themultiplewavelengthssuch that thephasechangeof thevarious wavelengths can be measured independently. With the current waveguidestructure it is required to image interferencepatternsondifferentpixel rowsof thecamera. Imaging the interference patterns on the same pixel rows would give anoverlap in spatial frequencies ( ij ijk d L ) for some wavelengths in combination

with variousdistances between the channels. For example, thephase changeof the457nmlaserincombinationwith100μmdistancebetweenthechannelscanhardlybedistinguishedfromaphasechangeofthe660nmlaserincombinationwiththe140


μmdistancebetween thechannels,because theybothhaveapproximately thesamespatialfrequencyatwhichthephasesignalshouldberedout.Inordertorealizetheseparationofthewavelengths,adiffractiongratingcanbeusedincombinationwithacoupleofachromaticlenses.ThethreeinterferencepatternsshouldbeimagedontoaCCDcamerawithachipheightof6.656mm(seesection3.6).Foroptimalspreadtheinterference patterns should approximately be imaged on 1/6, 3/6 and 5/6 on thecamera.Thismeansthatthedistancebetweentheouterinterferencepatternis2/3ofthe camera pixels, so 2/3 x 6.656mm= 4.437mm.The position y on the camera isdeterminedby the focal lens an thediffractiongrating that are used to split up theinterferencepatternsanddescribedby:

tan my f , (3.3)

where f is the focal length of the lens and the incident angle of the lens m can be

determinedfromthegratingformulaforincidentanglesof0º[8]:

sin ma m (3.4)

wherea isthedistancebetweenthecentresoftwoadjacentslits,mtheorderofthevariousprincipalmaximaand thewavelengthof the light.Assumingadiffractiongratingof300 lines/mm, 1 457nm and 2 660nm , 1m , the fcorresponding

to 4.437mmy isequalto69.8mm.

Nexttothedifferentwavelengths,differentpolarizationscanbecoupledintothewaveguidewhichalsohaveadifferentresponseonRIchangesintheevanescentfield.Therefore, theTMandTEmodesshouldbemeasured independently. ItwasnoticedthattheTEmodesofthe457nmlightdosometimesconverttoTMmodesonthechip.To prevent these modes interfering with themeasured TEmodes, a polarizer wasplaced between camera and waveguide. Alternatively, TE and TM modes can bemeasuredbothbysplittingthemupusingaWollastonprism(seeChapter6).

3.5.2RealizationThelightcoupledoutofthechipisfirstcollimatedusingacylindricalachromaticlenswithafocallengthf=75mm(Thorlabs,ACY254‐075‐A).Next,thelightisdirectedtoavisible transmission grating of 300 lines/mm (Thorlabs, GT‐25‐03, 25mmx25mm).Subsequently, a second cylindrical achromatic lens with focal length f = 50 mm(Thorlabs, ACY254‐050‐A) is used to image the interference patterns on the CCDcamera. For a lens of f = 50 mm, the difference in x between the first orders of

1 457nm and 2 660nm is3.2mm.AschematicsideviewisshowninFig.3.7.A

Linear Polarizer with N‐BK7 Protective Windows (400‐700 nm) (Thorlabs,LPVISE100‐A)isusedtofilteroutTEorTMpolarizedlight.Lightisonlycollimatedinthe vertical direction using cylindrical lenses. In lateral direction no lenses are

58 Chapter3

requiredastheinterferencepatternismeasuredatadistance L a .Fig.3.8showsthethreeinterferencepatternsimagesontopofeachotherondifferentpixelsrowsofthecamera,illustratingthatwearenowabletoindependentlymeasurephasechangesofthethreeinterferencepatterns.Interferencepatternsareimagedonmultiplerowstoincreasethedynamicrangeoftheinterferencepatternstoreducethephasenoise(seesection3.6).

Fig. 3.7: A schematic side view of how tow lenses and a grating are used to image theinterferencepatternsofthethreedifferentwavelengthsseparatelyonaCCDcamera,wherethewhitepartrepresenttheoverlappedlightthreewavelengths.Imageisnottoscale.

Fig.3.8: Typical camera image illustrating threemeasured interference patterns from threedifferentwavelengths,660nm,561nmand457nm,fromtoptillbottom.

3.6Detection

3.6.1RequirementsSimulations were done to determine which camera settings are important toaccuratelydetectaphasechangefroman interferencepattern.First,an interferencepattern ( 550nm ) based on 4point sources is generated consisting of a certainnumberofpixelsandwithacertaindynamicrange.Independentofanycamera,there


willalwaysbephotonshotnoiseandthereforeshotnoise(Poissonnoise)isaddedtothe interference pattern. No phase change is induced in the interference pattern,which means the determined phase should be zero. Consequently, the determinedphaseisequaltotheerrorinphase.Theerrorinthephaseisdeterminedasafunctionofthedynamicrangeofthecameraandthenumberofpixelsofthecamera.Foreachdynamic range and number of pixels an interference pattern was generated 1000times and a phase errorwas determined. The standard deviationof these values istakenasafinalmeasureforerrorinphase.Theresultsshowsthattheerrorinphaseshowsalinearrelationshipwiththedynamicrangeandnumberofpixelsonalog‐logscale (see Fig. 3.9). The slope is given by approximately ‐½ ,whichmeans that thenoisecanbedecreasedbymeasuringmorecounts(higherdynamicrangeandnumberofpixels)andthatthedecreaseofnoiseislimitedbythephotonshotnoise( 0.5~ N ).Increasingthenumberofpixelshasthesameeffectasincreasingthedynamicrange.Torealizeanerrorinthedeterminedphaseconsiderablylowerthan10‐4fringes,suchthat thecamera isnot the limiting factor, acamerawithadynamic rangeof16bits(65536) combined with a number of pixels of 4096 is required. Alternatively, thedynamic range can artificially be increased by imaging the interference pattern onmultiplepixelrowsandthenaddingupthesepixelrows.Aswasstatedbeforeitisalsorequired to have a camera with multiple pixel rows onto which the variousinterferencepatternscanbeimaged.

Fig.3.9:Standarddeviationofthedeterminedphaseof1000generatedinterferencepatterns(includingshotnoise)asafunctionofthenumberofpixelsandthedynamicrangeofacamera.

103 104 10510‐5

10‐4

10‐3

10‐2N

pixel=256

Npixel

=512

Npixel

=1024

Npixel

=2048

Npixel

=4096

Linearfit

fringes)

Dynamicrange

Equation y=a+b*xWeight NoWeighting

Adj.R‐Square

0.9998 0.99985 0.99995 0.9998 0.99994

ValueSlope ‐0.49739Slope ‐0.50197Slope ‐0.49902Slope ‐0.50432Slope ‐0.5013

60 Chapter3

3.6.2RealizationBecauseofitshighdynamicrangeincombinationwithacceptablenumberofpixelsinwidthandheight,anAltaU30‐OECCDcameraisused.Specificationsofthecameraareshown inTable3.2.At least10rowsof1024pixelswithadynamic rangeof12489wereselectedperinterferencepatternsuchthatthephasenoiseismuchlowerthantherequired10‐4fringes.

Table3.2:CCDcameraspecifications[9]CCDchip E2VCCD30‐11ArraySize(pixels) 1024x256PixelSize 26μmx26μmLinearFullWell(typical) 271KelectronsRMSNoise 21.6electronsDynamicRange(SNR) 12489(13bits=8192,14bits=16384)DigitalResolution 16bitsGain 4.05±0.01electronspercount(measured)Biaslevel 1236(measured)120secondDarkcurrent 1244counts(measured)Darkcurrent 0.1494countsperpixelpersecond(measured)Testtemperature ‐30ºC,Δ=51ºCExposureTime 30millisecondsto183minutes

3.6.3Characterization3.6.3.1ShutterThe Alta U30‐OE charge‐coupled device (CCD) camera has a mechanical shutter toblock incoming lightwhen reading out the pixel rows. Phase stability tests showedthat theusing thecamerashutter leads toaphasenoise in theorderof10‐3 fringeswhich ishigher than the required10‐4 fringes.Thisphasenoisemightbecausedbymovementofthecameraduetothismechanicalshutter.Maximally80fringesfallontoacamerawidthof26μmx1024=2.6624cm,whichmeansthat1fringecorrespondsto0.033cm.Consequently,aphasechangeof10‐3fringescorrespondstoamovementof330nm.Fig.3.10showsthephasechangecanbereducedsignificantlytoaphasenoise of <10‐4 fringes when not using the shutter of the camera. However, themechanical shutter is required to block the light during readout to prevent verticalsmear. Fig. 3.11 shows a camera image when no shutter is used, illustrating thevertical smear which makes it impossible to measure the interference patternsindependently.Thelaserscannotbeusedoptimallywhenmodulatingitbyturningitonandoff,sothelightisblockedduringthereadoutofthecamerausingamechanicalshutter (Uniblitz, LS2ZM2, Vincent Associates, Rochester, New York), driven by aVCM‐D1shutterdriver (Uniblitz) topreventvertical smearon theCCDcamera.Theshutter is placed on a position where all the wavelengths overlap and the shutterdriver is driven by a Arduino Uno boardwhich also drives the readout of the CCD


camera.Using the external shutter the phase change on a short time scale is lowerthantherequiredphaseprecisionof10‐4fringes.

Fig. 3.10: Phase change over time for a stability measurement with a shutter on (red) andwithoutusingashutter(black).

Fig.3.11: An example of a camera image of three interference patterns using a) the cameramechanicalshutterandb)noshutter,resultinginverticalsmear.

0 500 1000 1500 2000

‐5

0

5

10

15


Time(s)

ShutteronNoshutter(shutteropen)

x10‐3

62 Chapter3

3.5.3.2CCDCameratemperatureTo determine the influence of the camera cooler, the phase stabilitywasmeasuredwith thecoolerand fanoffandon.Turningoff thecoolerof thecameraresults inaveryunstabletemperature,resultinginanon‐stablephasesignalasseeninFig.3.12a/b.Fig3.12c/dshowthetemperatureandcorrespondingphasestabilitywhenthefan was turned off. When turning the fan off, the temperature starts increasingresulting in a phase drift, probably caused by expansion of materials, resulting inmovement of the CCD chip. As expected, the phase change is the highest for thehighest spatial frequency, which has the highest number of fringes on the chip.Therefore,onefringeisthesmallestatthehighestspatialfrequencyandconsequentlyamovementoftheCCDchipresultsinthehighestphasechangeforthisspatial

Fig.3.12:a)TemperatureofCCDcameraandb)correspondingphasechangeovertimeforthe660nm laserwhen turning thecamera fanandcooleronandoff, andc) temperatureofCCDcameraandd) correspondingphase changeover time for the660nm laserwhen turningoffonlythecamerafan.

0 5000 10000 15000‐30

‐20

‐10

0

10

20

30

0 5000 10000 15000‐4

‐2

0

2

0 200 400 600 800 1000‐30

‐20

‐10

0

10

20

30

0 200 400 600 800 1000‐2

0

2

4c d

b

fan+cooleron

Tccd(°C)

Time(s)

fan+cooleroff

a x10‐2


Time(s)

redd12redd23redd34redd13redd24redd14

Tccd(°C)

Time(s)

x10‐2


Time(s)

redd12redd23redd13redd34redd24redd14

fanoff


frequency.Tokeepthephasechangestable,thetemperatureofthecamerashouldbekept stable and this was done during the measurements by setting cooler of thecameraat‐30ºCandthefanoninthehighestmode(dataofdifferentfanmodesisnotshown,butbestresultswereobtainedbyhighestfanmode).

3.5.3.3MeasuredcameranoiseThenoiseofthecamerawascharacterizedusingaphotontransfercurve(PTC)aswasalsoillustratedby[10,11].TheCCDcameracanbeseenasasystemblockwithlightasinputanddigitaldataasoutput.Shotnoiseduetothenatureofphotonsistheonlynoiseattheinput.SoanydifferencebetweenthenoiseattheinputandoutputmustbecausedbytheCCDcamera.TheCCDcamerawasexposedwithanilluminationfieldusing a Farnell 1814434 green light‐emittingdiode (LED) (525nm). For a rangeofintensities (~1260‐64000 counts), two consecutive frames were captured andsubtractedtoremoveframe‐to‐frameoffsetsandphoto‐responsenon‐uniformity.Therootmeansquare(RMS)noiseofthecameraiscalculatedbytakingthesquarerootofthevarianceandisgivenby[10]:

2

1

1 1 2 2

2

pN

i ii

p

X M X M

N

, (3.5)

whereNpisthenumberofpixels,X1itheindividualpixelvaluesofthefirstframe,X2ithe individualpixelvaluesof thesecond frame,M1 themeanofall thepixelsof thefirstframeandM2themeanofall thepixelsofthesecondframe.Becausethenoisewasinitiallydoubledbysubtractionofthetwoframes,thereisafactoroftwointhedivisor.APTCcurveisnowgeneratedat CCDT ≈‐30ºCbyplottingtheRMSnoiseasa

function of themean signal in a log‐log plot which is shown in Fig. 3.13. Fig. 3.13shows that for a number of counts larger than 1000 the shot noise is the limitingfactorastheslopeofthefigureisequalto0.5.Shotnoiseisequaltothesquarerootofthenumberofcountswhichappearsasa0.5slope ina log‐logplot.Toconfirmthatthiscameracanbeusedtomeasureaphaseprecisionof10‐4fringesaninterferencepatternbasedon4pointsources(representinga4channelwaveguide)was imagedonto the CCD by an achromatic lens using a 525 nm Farnell 1814434 LED source,settingthemaximumnumberofcountsmeasuredbytheCCDcamerato≈60000bytuningtheintensityoftheLED.Withanexposuretimeof0.03sandthe CCDT ≈‐30ºC,

theinterferencepatternsweremeasuredoveramultiplenumberofrows rowsN (4,16,

50, 100 and 200 rows) which were added up. The phase change over time wasdeterminedoftheseselectedinterferencepatternsandfor rowsN =100plottedinFig.

3.14a.Todeterminetheshorttimescalenoiseandexcludinglongtimescale(thermal)drift,thephasechangewassmoothedwithaN=65linearSavitsky‐Golayfilter.This

64 Chapter3

signalwassubtracted fromtheoriginallymeasuredsignal (seeFig.3.14b)and fromthe difference (Fig. 3.14c) the noise was defined by determining its standarddeviation.Thephasenoisewasaveragedforallspatialfrequenciesanddeterminedasafunctionof rowsN andcomparedwithsimulateddatawhichisshotnoiselimited.For

spatial frequency d12 which belongs to the distance between channel 1 and 2, themeasured phase noise is close to the simulated shot noise limited case. For largerspatialfrequencies,thenoiseincreasesasshowninFig.3.14a.Thismightindicatethatimaged interferencepatternvibrateswith respect to the camera, because the lowerspatial frequenciesare lessaffectedbecauseoftheir lowernumberof fringesonthecamera and therefore a larger size per fringe. The measured phase noise

correspondingtod14was≈ 55 10 whichcorrespondstoavibrationof 5nm (ford14 approximately 65 fringes fall onto the 6.656mmwhichmeans that 1 fringes isequal to100μm).Theabsolutephasedriftona large timescalealsoscales linearlywith the spatial frequencies which indicates that the interference pattern imageslightlydrifts(<10‐3fringes≈416nmover≈4000s)withrespecttothecamera.Asaconclusion we can say that the position of the camera with respect to the camera

shouldbeverystable( 10nm foraphasenoiseof 410 ),butifphasedriftornoiseis causedbymovement itwill scale linearlywith themeasuredphasenoises at thespatial frequencies belonging to different distances between the YI channels.Moreover,itcanbeconfirmedthatthecameranoiseonashorttimescaleisshotnoiselimitedandmeasuringtheinterferencepatternsovermorethan10pixels,thephasenoiseismuchlowerthantherequired10‐4fringes.

Fig 3.13: Photon transfer curve showing the RMS noise as a function of the mean signalmeasured for the CCD camera measured at TCCD = ‐30 ºC. The red line has a slope of 0.5indicating that the measured points fit the shot noise. The blue line with slope of zerorepresentsthereadnoiselevel.

10‐2 10‐1 100 101 102 103 104 105

100

101

102

103

(counts)

Mean(counts)


Fig.3.14:a)MeasuredphasechangeovertimewheninterferencepatternwasimagedontoCCDcamera for all six spatial frequencies andafter summationof100 rowsof the camera, b) thephasechangeovertimefor rowsN =200andspatialfrequency(sf)d12,smoothedwithaN=65

Savitsky‐Golayfilter,c)thesubtractedsignalofthemeasuredandsmoothedsignalford12and

rowsN =50,100and200,andd)thephasenoise(standarddeviation)asafunctionof rowsN of

themeasureddata(averageofallsf’sandd12only)andsimulateddata(averageofallsf’s).

3.7Dataprocessing

3.7.1RequirementsThe phase change should be detected as accurately as possible from the collectedinterference patterns with the CCD camera. First, an FFT should be applied to theinterference patterns. Next, the phase change should be red out at the spatialfrequenciesbelongingtothewavelengths,cameradistanceandchanneldistances.Thephaseerror(PE=differencebetweenexpectedinducedphasechangeandmeasuredphasechange)andcrosstalk(CT=deviationwherenophasechangeexpected)should

0 1000 2000 3000 4000‐12

‐8

‐4

0

0 1000 2000 3000 4000‐12

‐8

‐4

0

0 1000 2000 3000 4000‐1.0

‐0.5

0.0

0.5

1.0

100 101 102

10‐5

10‐4

10‐3

dc

bNrows=100


Time(s)

d12d23d34d13d24d14

ax10‐4 x10‐4


Time(s)

Measured,Nrows=200,d12Smoothed,Nfilter=65

x10‐4 d12


Time(s)

Nrows=50

Nrows=100

Nrows=200

SimulatedMeasured,Nfilter=65,allchannels

Measured,Nfilter=65,channeld12

Nrows

66 Chapter3

beminimized.PTandCTcanbereducedbyanartificialincreaseofspatialfrequencyresolutionandapplicationofasufficientwindowing[1].Thespatial frequenciescanbereducedbya factorof~2usinganartificial increaseof thespatial frequency[1].The spatial frequency resolution cannotbe increased indefinitely,because theerrorintroducedbyadecreaseofthepeakamplitudes,whichisalsoaresultoftheartificialincrease, will dominate the effect of spatial frequency increase. Using a Hanningwindow,thePEandCTcanalsobereducedbyafactorof~5[1].

3.7.2RealizationToreadouttheinterferencepattern,todeterminethephasechangeovertimeandtoapply the above named techniques a Labview programwaswritten. First, the CCDchipisredoutasa 256 1024 (rows columns)matrixwithindividualpixelvalues.Next, a selected number of rows of this matrix is added up separately for eachinterferencepatternofeachwavelength.Alsothenextstepsaredoneseparately foreach wavelength. Subsequently, a window is applied on the summed interferencepatternsafterwhich1024zerosareaddedtodecreasethePEandCT.Then,anFFTisapplied after which the phase signal is read out at a selected spatial frequencyamplitude corresponding to a certain channel combination and wavelength. Phasejumpsof 2 arecorrectedusingaself‐writtenunwrapfunction.Moreover,thespatialfrequencyamplitudeandcameratemperatureTCCDarerecorded.

3.7.3CharacterizationComputer generated images of interference patterns (based on our four channelYoung interferometer) were used to verify the correct working of the Labviewprogram which is used to collect interference patterns and determine thecorresponding phase changes during an experiment, to investigate artefacts whendeterminingaphasechangefromafiniteinterferencepatternandtotestwindowingfunctionstoreduceartefacts.Calculationsofthegeneratedinterferencepatternsarebasedonfourpointsourcesdistancedat60μm,80μmand100μm(seeFig.3.2),animageddistanceof17.5cm,andwavelengthsof457nmand660nm.Thegeneratedimages have awidth of 1024 pixels, a height of 100 pixels and a dynamic range of60000countsandincludeshotnoise(Poissonnoise).Tosimulatearealexperiment,a

timevarying effN (~ )isintroduced(seeFig.3.15a).Foreachcorrespondingtime

pointanimageiscalculatedandstoredforlateranalysis.Next,theLabviewprogramloads the images, adds up the selected number of rows on which the interferencepatternsareimagedandsubsequentlyadds1024zerostoincreasespatialfrequencyresolution to improve phase readout. Optionally, a windowing function is appliedbeforeaFastFourierTransform(FFT)isexecuted.Fromtheresultingspectrum,thephaseandamplitudeofthesixdifferentspatialfrequenciesaredeterminedofwhichwenowanalyseonepeakbelongingtoonechannelpair.Fig.3.15ashowsanexample


of the induced and calculated effN for the twowavelengths. The same figure also

showsthedifferencein effN betweentheinducedanddeterminedvalues.Fig.3.15b

showsthedeterminedamplitudeandFig.3.15cshowsthe sn determinedwiththe

ratio‐basedapproach(seeCh.4,page79forexplanationofratio‐basedapproachand

sn )basedontheinduced(ind)anddetermined(det) effN .Fig.3.15d,eandfshow

the effN , the amplitude and the determined sn of a real measurement. Similar

fluctuations(whichwecallartefacts)areseeninintheamplitudeand sn whenno

windowingisappliedonthesimulatedinterferencepatterns,indicatingthatboundaryeffectsoftheshiftinginterferencepatternarecausingthemeasuredartefacts.These

artefactsaremostprominentduring(fast)changesintheinduced effN .

Fig.3.15:Comparisonoffluctuationsinthe effN asdeterminedbyusingsimulateddata(a,b,c)

andmeasureddata (d,e,f) in a typicalmeasurement composedof abulk signal followedby acombinationofabulksignalandasignalarisingfromsurfacebinding.a) effN induced(ind)

and determined by analysing simulated data without applying a window (det) and thedifference between detected and induced effN , b) the amplitude determined by the FFT

corresponding to the from which effN is determined, c) n as induced and as

determined using the ratio‐based approach, d) the effN of a measurement where first D‐

glucose and later D‐glucose and 85 nm beads are added to the sensor surface, e) thecorresponding amplitude of the FFT‐spectrum at the measured spatial frequency and f) thedetermined n duetobindingofthebeadsandbulkchangesduetotheD‐glucose.

0 400 800

‐2

0

2

4

0 400 80018.0

18.5

19.0

19.5

20.0

0 400 800

0

2

4

6

8

0 1000 2000

0

1

2

3

4

0 1000 20007.6

7.8

8.0

8.2

0 1000 2000

0

2

4

6

8

Neff

Time(s)

660,ind457,ind660,det457,det

‐1

0

1

2

3

4x10‐6

660,det‐ind457,det‐ind

x10‐4

Neff

Amplitude(counts)

Time(s)

457nm660nm

n u

nscaled

Time(s)

Binding,indBulk,indBinding,detBulk,det

x10‐5

ed f

cb

x10‐4

Neff

Time(s)

660nm457nm

a

Amp 6

60nm(counts)

Time(s)

660nm 1.6

1.8

2.0

2.2

x107

x107x107

457nm Amp 4

57nm(counts) x10‐5

n u

nscaled

Time(s)

85nmbeadsD‐glucose

68 Chapter3

To investigate this further, a linear (~ effN )was inducedand thepeak‐to‐

peak(pk‐pk)valuesoftheartefactsin effN weredeterminedtotesttheinfluenceof

theappliedwindowonthesizeoftheartefacts(bothinamplitudeand (~ effN )).

Fig.3.16ashowsthelinear effN determinedfromthelinear ,andthedifference

between thedeterminedand induced effN (= effN error, abbreviatedasNE in the

remainder of the text) for the case inwhich nowindow is applied and the case inwhich a in Labviewbuild‐inBlackman‐Nuttall window is applied. Fig. 3.16b showsthemeanpk‐pkvaluesoftheNEofdifferentchannelcombinations,determinedfrom measuredatthespatialfrequencies(sf)belongingtothechannelcombinationin

which a n was generated in one of the channels as well as from the channelcombinationsinwhichno n occurred(cross‐talk,CT).Todemonstratetheeffectofwindowing,resultsareplottedfordifferenttypeofwindows.Forexample,applyingaBlackman‐NuttalloraDolph‐Chebyshevwindow,bothCTandNEcanbecompletelysuppresseddown to the shot‐noise level as theartefactswerenotvisibleabove theshot‐noiselevel.

Next, the use of differentwindows on real experimental data is evaluated. Themeanpk‐pkvaluesofCTaresimilar to themeanpk‐pkvaluesofNE.Therefore, themeanpk‐pkvaluesofCTareusedasameasureforthepk‐pkvaluesofNEfortherealexperiments,asitisnotpossibletodeterminethefluctuationsinthemeasuredsignal

of effN ,becausetheslopeofthemeasuredsignalismuchhigherthantheslopeofthe

artefacts.Alsonoinduced effN canbedeterminedandconsequentlyitisnotfeasible

todeterminethedifferencebetweenameasuredandinduced effN todeterminethe

NE.Rawdata(interferencepatterns)weresavedanddifferentwindowswereapplied

on the measured data. Fig. 3.16c shows the fluctuations in effN due to CT for a

typicalmeasurementwhere firstD‐glucose only (between t≈500 and1000 s) andsubsequentlyD‐glucoseplus85nmbeads(betweent≈1500and2000s)wereaddedsimultaneouslytothesensor.Fig3.16dshowsthepk‐pkvalueofthefirstartefactin

effN (at t ≈500 ‐ 700 s)due toCTasa functionof the appliedwindow.Applying

differentwindowsdidnotleadtosignificantdifferencesinperformance.Probablythemeasuredinterferencepatternsalreadycontainsahardware‐inducedwindowing.Thenon‐idealpropertiesofthelenses,thegratingandbeamshapecauseanellipse‐shapedimageoftheinterferencepatternasshowninFig.3.8.BeforeapplyinganFFTonthemeasured patterns, pixel intensities of each column in the recorded image aresummedandthereforetheellipse‐shapedimageresultsinaninterferencepatterninwhichtheintensitiesaresuppressedattheedgeswhichmeansthatitalreadyconsistsofacertainwindow.Furthermore,itwasseenthatdifferentopticsresultedinchangeinartefacts.Therefore,itshouldbetestediflargerlensesandgratingcanreducethe


artefactsinthesignalsof effN .TheanalysisofanexperimentwithatwochannelYI

(two channels were blocked with a mask) showed similar artefacts as for a fourchannelYI,whichmeanstheartefactsare in themeasuredsignal itselfandnotonlyinducedbyCTas there isonlyonespatial frequencypeakfora two‐channelYI.TheamplitudeofthemeasuredoscillationsinCTareintheorderof10‐2fringes.Inducedsignals should be significantly larger than the artefacts. The artefacts influence thefinalamplitudeofthesignalwithamaximumdeviationequaltotheamplitudeoftheoscillation.

Fig.3.16:Theeffectofapplyingdifferentwindowsonthedataontheerrorinthedeterminedphasechange(~ effN ).a) effN inducedandthedifferencebetweendeterminedandinduced

effN (NE) applying no window (magenta) and a Blackman‐Nuttall window (green), b) the

peak‐to‐peak(pk‐pk)valuesofthefluctuationsin effN (forbothNEandCT)forthedifferent

appliedwindow,c)themeasured effN ofameasurementchannelandtheCTinasidechannel

and d) the pk‐pk values of the fluctuations in effN due to CT as a function of the applied

windowandselectedpixelsoftheinterferencepattern.

0 100 200 300‐4

‐2

0

2

4

6

1 2 3 4 5 610‐8

10‐7

10‐6

10‐5

0 500 1000 1500 2000 2500

‐2

0

2

4

1 2 3 4 5 62.0

2.5

3.0

3.5

4.0

4.5

5.0

Neff

Time(s)

Induced

‐1

0

1

2

3

4x10‐4

NowindowBlackman‐Nuttall

x10‐6

Neff

CTNE

(Neff){Pk‐pk}

Windowingfunction

W1=nowindowW2=HammingW3=Gaussian,=0.2W4=HanningW5=Blackman‐NuttallW6=Dolph‐Chebyshev,100dB

x10‐4

Neff

Time(s)

660,meas.sf.457,meas.sf.

pk‐pk

0

2

4

6660,sidesf.457,sidesf.

Neff

x10‐6

x10‐6c d

b

Measured

W1=nowindowW2=HammingW3=HanningW4=Gaussian,=0.2W5=Blackman‐Nuttall(B‐N)W6=Dolph‐Chebyshev,100dB

Neff{Pk‐pk}

Windowingfunction

Simulateda

70 Chapter3

3.8Overviewsetup

An overview of the setup is given in Fig. 3.17. The setup is coveredwith Thorlabsenclosure accessories (see Fig. 3.17a) to protect it from air currents, dust and lightinterfering with the measurements. It was measured that air current can stronglyinfluence the phase signal as shown in Fig. 3.18 which also shows that the laserscoveringboxshouldstayclosedduringmeasurements.Fig.3.17b/cshowanoverview

Fig.3.17:a)SetupcoveredwithThorlabsenclosureaccessoriestoprotectinfromaircurrent,dustandlightinterferingwiththemeasurements,b/c)thesetupwhenusingend‐firecouplingandd/e)thesetupwhenusingfiberbutt‐endcouplingandf)the660nmlightcoupledoutthe4channelwaveguidestructure.


of the setupused for end‐fire coupling. Fig. 3.17d/e showanoverviewof the setupusedforfiberbutt‐endcoupling.Fig.3.17fshows4pointsourcesofthe660nmlightcoupled out of the waveguide structure which correspond to the output of the 4channelsoftheYI.

Fig.3.18:Phasechangeovertimewhileturningfanonandoffandopeningandclosingofthelasercoveringbox.

3.9Characterizationofphasenoiseanddrift

Thephasenoise(shorttimescaleofseconds)andthephasedrift(noiseonlongtimescaleofhundredsofseconds)weremeasuredforbothincouplingmethodsasshownin Fig. 3.19. The determined phase noise on a short time scalewas determined bysmoothing thephasewithaN=15Gavitsky‐Golay smoothing filterand subtractingthis from the original phase signal and by taking the standard deviation of thissubtractedsignal.Thisleadtoadeterminedaveragephasenoiseofthemeasurementsof2x10‐5and6x10‐5 fringes forbutt‐endcouplingandend‐firecouplingrespectivelywhichistypicalforallthemeasurements.Itcanbeconcludedthatthissufficientasitislowerthantherequired10‐4fringes.

The usual phase drift is smaller than 5x10‐3 fringes per 1000s for the end‐firecoupling and 1.5x10‐2 fringes per 1000s for the butt‐end coupling. The higher driftwith the butt‐end coupling might be caused by the ULTRAlign™ Precision XYZPositioning Stage and the long fiber tipwhichmight be not perfectly stable. Itwasseen that after stabilising the stage for a couple of hours the drift was reduced tovalues comparable with the end‐fire coupling. It should be noticed that there is adifferencebetweenthephasedriftofthedifferentspatialfrequencies.

An ideawas touse the correlationbetween themeasuredphase changesof thedifferentwavelengthsindifferentchannelsasillustratedinFig.3.20.Thephasenoiseof thethreewavelengths ispresentedforall thesixspatial frequenciesbelongingto

0 400 800 1200

‐0.6

‐0.3

0.0

0.3

0.6

660nm,d24

boxopenfanoff

fanoff

fanon


Time(s)

660nm,d13

x10‐2

fanon

72 Chapter3

the distances between the four channels. The Pearson product‐moment correlationcoefficientsweredeterminedforallthephasesignalsandvariedfrom0.441( 34d for

457 nm and 660 nm) to 0.998 ( 24d for 561 nm and 660 nm), where a correlation

coefficient of 1 is fully correlated). This means that there is a correlation betweenthesephasesignals.However,thiscorrelationisnotconstant.Sometimes,thedriftinthe660nmlaseristhehighest,butalsosometimesthe561nmorthe457nmlaser.No clear trend between the phase signal of the wavelengths could be observed.Therefore,thewavelengthscouldnoteasilybeusedtocanceloutphasedrifts.Lateralmovementofthecamerawithrespecttothechipcanleadtophasedrift,butthiscanclearlybedetectedasitscalesinverselywiththespatialfrequencies.ItisalsoknownthattemperatureplaysanimportantroleasitcanleadexpandingmaterialsandstressinmaterialswhichcanleadtoRIchangesandchangesindistances.ChannelsintheYIwaveguidewereplacedclosetogethertohavearelativelyconstanttemperatureoverthechip.Tohaveabetterheatsinkforthechiptheplasticchipholderwasreplacedbyabrasschipholder,butthisdidnotleadtoanyreductionofthephasedrift.Moreover,phasemeasurementswithout awaveguide butwith two slits, but this did also notgaveinsideintheoriginofthedrift.Thismeansthatthemeasurementspresentedinthenextchaptercontaindrift.SimulationsinChapter4showthatlineardriftresultsin a linear drift in the signal of n . The drift in the measurements should bedeterminedbydetermining thebaseline. Ifmeasurements showa linearphasedriftthedrift canbe compensated for by fitting it linearly and subtracting it. Preferably,inducedsignalsshouldbesignificantlyhigherthanthephasedrift.

Fig. 3.19: a) Phase change of 660 nm laser measured for fiber incoupling and free spaceincoupling for two measurements per incoupling methods measured each at six spatialfrequenciesbelongingtothesixchannelcombinationoftheYIandb)azoom‐inofftwosignalsofa).

0 200 400 600 800 1000

‐20

‐15

‐10

‐5

0

5

10

200 250 300 350 400‐10

‐8

‐6

‐4

‐2

0


Time(s)

FiberincouplingFreespaceincoupling

x10‐3 b


Time(s)

FiberincouplingFreespaceincoupling

x10‐4a


Fig.3.20:Typicalphasenoiseovertimeof457,561and660nmlasercoupledintowaveguideusingfiberbutt‐endcouplingforsf’sbelongingtodistancesbetweenchannelsa) 12d ,b) 23d ,c)

13d ,d) 34d ,e) 24d andf) 14d .

0 500 1000 1500 2000‐2.0

‐1.5

‐1.0

‐0.5

0.0

0.5

0 500 1000 1500 2000‐0.5

0.0

0.5

1.0

1.5

2.0

0 500 1000 1500 2000‐2.0

‐1.5

‐1.0

‐0.5

0.0

0.5

0 500 1000 1500 20000.0

0.5

1.0

1.5

2.0

2.5

0 500 1000 1500 2000‐0.5

0.0

0.5

1.0

1.5

2.0

0 500 1000 1500 2000‐2.0

‐1.5

‐1.0

‐0.5

0.0

0.5

e f

c d

bd12


Time(s)

660nm561nm457nm

x10‐2 d23x10‐2


Time(s)

660nm561nm457nm

d13x10‐2


Time(s)

660nm561nm457nm

d34x10‐2


Time(s)

660nm561nm457nm

a

d24x10‐2


Time(s)

660nm561nm457nm

x10‐2


Time(s)

660nm561nm457nm

d14

74 Chapter3

Acknowledgements

IwouldliketothankKeesvanderWerfandRobertMolenaarfortheirhelpinthelab;Robert especially for the help with the shutter, camera and Arduino and Keesespecially for the help with finding the origin and the characterization of theoscillationsduetothereflectionsbetweenfiberandwaveguide.Furthermore,IwouldliketothankErwinHondebrinkforthehelpwithLabviewprogrammingandthankstoMartijnStopelforthehelpwiththedesignofthesetup.Finally,IwouldliketothankXiOPhotonicsB.V.andinparticularRonaldDekkerforthehelpwiththerealisationofthefiberbutt‐endcoupling.

Appendix3.AFocalpositionusinga4flenssystem

Assumingsmallanglesandsmalldistancesrelativelytotheopticalaxisofthesystem(paraxialapproximation),theangleandpositionatacertainpointcanbedeterminedwithRaytransfermatrixanalysis.FortheRaytransfermatrixanalysis,athinlenscanbedescribedbythefollowingmatrix:

1 0

11

f

, (3.A.1)

wherefisthefocallengthofthelens.Adistancedcanbedescribedbythefollowingmatrix,assumingpropagationinair(n=1):

1

0 1

d

. (3.A.2)

The 4f lens systemwhich is used to focus the light into thewaveguide structure isshowninFig.3.A.1andconsistsof lensesanddistancesbetweenthe lensesandcantherefore be described using equations 3.A.1 and 3.A.2 in reverse order of thepropagation of the light trough these components. The position and angle at f aredeterminedusingthefollowingequation:

03 2 1 2 1

03 2 1

1 0 1 0 1 01 1 1 1

1 1 11 1 10 1 0 1 0 1 0 1

f

f

y yf f f f f

f f f

, (3.A.3)


where 1f , 2f and 3f arethefocaldistancesoflens 1L , 2L and 3L respectivelyand 0y

and 0 are the position and angle at the Gimbal mirror. Equation 3.A.3 can be

simplifiedto:

1 30

2

2 0 1 0

1 3 2

f

f

f fy f

f y ff f f

. (3.A.4)

Thismeansthethatposition fy isdeterminedbytheangle 0 iny‐directionatthe

Gimbalmirror and the focal distances of the used lenses. In a similarway fx is a

functionoftheangle 0 inx‐direction.

Fig.3.A.1:Schematicoverviewof4flenssystem.

References

1. A.Ymeti,“DevelopmentofamultichannelintegratedYounginterferometerimmunosensor,”UniversityofTwente(2004).

2. M.Hammer,“1‐Dmultilayerwaveguidemodesolvereffectiveindexapproximation”(2015),retrieved23‐07‐2015,http://www.computational‐photonics.eu/oms.html.

3. CoboltAB,“OwnersManualModel05‐01”(CoboltAB,February2015,2015),retrieved01‐06‐2015,http://www.cobolt.se/wp‐content/uploads/2015/03/Owners‐Manual‐05‐01_150206‐1.22.pdf.

4. CoboltAB,“OwnersManualModel04‐01”(CoboltAB,March2015,2015),retrieved01‐06‐2015,http://www.cobolt.se/wp‐content/uploads/2014/10/Owners‐Manual‐04‐01_v1.63_20150305.pdf.

5. P.Kozma,F.Kehl,E.Ehrentreich‐Förster,C.Stamm,andF.F.Bier,“Integratedplanaropticalwaveguideinterferometerbiosensors:Acomparativereview,”Biosens.Bioelectron.58,287‐307(2014).

6. “IntegratedLaserBeamCombiner”(XiOPhotonicsB.V.,2013),retrievedhttp://www.xiophotonics.com/index.php/products/49‐integrated‐laser‐beam‐combiner.

76 Chapter3

7. “Polarization‐MaintainingFiber:PandaStyle”(Thorlabs,01‐10‐2013),retrieved2015‐09‐10,http://www.thorlabs.us/Thorcat/13900/PM460‐HP‐SpecSheet.pdf.

8. E.Hecht,“Optics,”4ed.(AddisonWesley,2002),p.477.9. G.Xiong,“CameraTestReport.”10. D.Gardner,“Characterizingdigitalcameraswiththephotontransfercurve”(Summit

Imaging),retrieved16‐07‐2015,http://www.couriertronics.com/docs/notes/cameras_application_notes/Photon_Transfer_Curve_Charactrization_Method.pdf.

11. P.Lytaev,A.Hipp,L.Lottermoser,J.Herzen,I.Greving,I.Khokhriakov,S.Meyer‐Loges,J.Plewka,J.Burmester,M.Caselle,M.Vogelgesang,S.Chilingaryan,A.Kopmann,M.Balzer,A.Schreyer,andF.Beckmann,“CharacterizationoftheCCDandCMOScamerasforgrating‐basedphase‐contrasttomography,”inDevelopmentsinX‐RayTomographyIX,(Proc.ofSPIE,2014),921218.

Chapter41

Differentanalysisapproachesforsize‐selectiveanalytedetectionAbstractThis chapterpresents a description andadetailed studyof different signal analysisapproachesthatarerequiredtoobtainthesignalfromthedifferentsubstancesfromthe measured phase changes at the different wavelengths. It was found that atheoretical approach as presented in Chapter 2 is exact but in practice difficult toimplementbecauseofthemanyparametersthathavetobetuned.Therefore,wealsodevelopedamuchmorepracticalratio‐basedapproachbasedontheratiosof effN ’s

measuredatdifferentwavelengths.Theseratiosweredeterminedindependently for85nm carboxylatedpolystyrenebeads (representing specific binding of e.g. viruseswhich have approximately this size), protein A (representing non‐specific binding)and D‐glucose (representing bulk changes). These ratios are used to discriminatebetween the n ’s induced by these substances which are applied in the proof‐of‐principleexperimentsforsize‐selectivedetection.Ontheotherhand,usingthisratio‐basedapproachitisnotpossibletodetermineanabsolutevalueof n .However,thisis not strictly required, because calibration measurements, which are usually alsocarriedoutbyasinglewavelengthYI,canbeusedtocalibratethemeasuredvaluesofn with an analyte concentration. The theoretical and ratio‐based approach were

combinedtodeterminetheabsolutevalueof n andthecorrespondingsurfacemasscoverage.Moreover, theapproacheswerecombinedtodeterminethe thicknessesof85nmbeads tobe≈72.6nmandproteinAtobe≈2.1nmwhichare inagreementwithliterature.Furthermore,theinfluenceoftheparametersthatarerequiredfortheanalysis approaches (e.g. waveguide core thickness, waveguide refractive indices,ratiosof effN ’s )on thedetermined n ’s caused indifferent layersorbydifferent

substanceswastested.Finally,simulationsofarealexperimentalsettingconfirmthetheoretical analysis given in Chapter 2 that noise, drift and artefacts show upenhancedinthedetermined n ’s.

Partofthischapterisusedinmanuscript:H.K.P.Mulder,C.Blum,V.Subramaniam,J.S.Kanger,“Size‐selectiveanalytedetectionwithaYounginterferometersensorusingmultiplewavelengths”,manuscriptinsubmission

78 Chapter4

4.1Introduction

Usually,aYIisusedtodetermineaneffectiverefractiveindexchange effN atasingle

wavelength. Consequently, an average refractive index change n is determined,based on any n in thewhole evanescent field. Determining the effN atmultiple

different wavelengths enables discrimination between multiple different n ’s ordetermineanalytethicknesses.Inthischapterwepresentthreeapproacheswhichcanbe used for this: a theoretical, a ratio‐based and a combined theoretical and ratio‐basedapproach(combinedapproach).

4.2Theoreticalanalysisapproach

Determining the effN at two different wavelengths, enables the discrimination

between n ’s originating from twoarbitrarily chosen layerswithin the evanescentfieldasshowninChapter2,usingthefollowingequation:

1

;l effn S N

, (4.1)

wherevector ln iscomposedoftheelements ,l jn whichistheRIchangeinthejth

layer, the coefficients of matrix S are given by sensitivity coefficients

, , ii j eff jS N n ,whicharethederivativesof effN oftheithwavelengthwithrespect

tonoccurringinthejthlayerandthevector effN iscomposedoftheelements ,eff iN

which are the effN measured at the ithwavelength. The sensitivity coefficients are

determinedbythestructureofthewaveguide(corethicknessandRIsofthevariouslayersofthewaveguide),thewavelengths,thepolarizationandtwoarbitrarilychosen

layers(withintheevanescentfield).The ln canbedeterminedbymeasuring effN

and multiplying this by the inverse of matrix S . A homogeneous ln per layer is

determined and therefore this method is less suitable for situations where n ’sinduced by different substances occur in the same layer. However, assuming a lowconcentrationofanalytes,forexamplebulkeffectscanbecancelledoutbysubtractionof n determinedfrommultiplelayerstofinallyarriveatananalyteconcentrationasshowninChapter2.Nevertheless,thismethodrequirestuningofmanyparameterstodetermine the correct values for the sensitivity coefficients that are required todetermine the correct value of n and Canalyte. Hence, this method is exact but inpracticedifficulttoimplementbecauseofthemanyparametersthathavetobetuned.

Differentanalysisapproachesforsize‐selectiveanalytedetection 79

4.3Ratio‐basedanalysisapproach

4.3.1TheoryWe also developed a much more practical ratio‐based analysis approach. Thisapproachusestheratiosof effN atmultiplewavelengths,causedby n ’soriginating

fromdifferent substances sn . Each substancewith a different sizewhich causes a

sn intheevanescentfieldresultsindifferent effN ’sforeachwavelengthbecauseof

the different electric field distributions of the respective wavelengths. When a

substancesmisaddedtothesensoritinducesa smn (menumeratesthesubstances,m

= 1, 2, 3 …), this results in an effective refractive index change , keffN at the

wavelength k (k=1,2,3…).Assumingthattheresponseofthesensorhasalinear

relationshipwiththeamountofsubstancesm thatisaddedtothesensor,thiscanbewrittenas:

,, s

s

k

k m

m

effeff

NN n

n

, (4.2)

where , sk meffN n isthesensitivitycoefficientofthesensorforaRIchangeinduced

bysm.Assumingthatsubstancesdonotinfluenceeachother’sresponseofthesensor,multiple substances inducing different ratios can be discriminated from each otherusing:

seff subN S n , (4.3)

where the vector effN is composed of the elementskeff,N , the vector sn is

composed of the elementsms

n , and the coefficients of matrix subS are given by

, , sk mk m effs N n .

Next,wedefinetheratiooftwo effN ’smeasuredattwowavelengths k and l ,inducedbyasubstancesmas:

, ,

, ,

,/

,

m m

eff eff m mk km

k l m m

eff eff m ml l

s ss s k ms

s sl ms s

N N n n sR

sN N n n

, (4.4)

80 Chapter4

whereweneglectthedispersioninms

n .Wecanexpresstheelementsofmatrix subS

in terms of the ratios /m

k l

sR . Due to the fact that / /1m m

k l l k

s sR R and

/ / /m m m

k l k j l j

s s sR R R wecanrewriteequation(4.3)as:

seffN R n , (4.5)

where isamatrixwiththeoreticalsensitivitycoefficientsofwhichtheelementsaregiven by , , s

k mk m effN k mn and , 0 k m k m , and the coefficients of

matrix R aregivenby , /m

k m

sk mr R .Todetermine sn equation(4.5)canberewritten

as:

1 1

s effn R N

.2 (4.6)

The experimental procedure is now as follows. First a characterization step is

requiredtodeterminematrix1

R

bymeasuringtheresponseofthesensorforeach

individualsubstance.Thematrix1

canberegardedasascalingmatrixthatallows

to get absolute values of sn . When the coefficients of this matrix are unknown,

, k m k m canbesetas1suchthat1

becomestheidentitymatrix,whichmeans

that sn is fullydeterminedby themeasuredratios /m

k m

sR .Without thescaling it is

possible to compare the unscaled smn with sm

n of a different measurement.

However, the amplitude of smn should not be compared with sn

n as scaling is

required.Nowthatweknowthematricesofequation(4.6),wecanusethisequationforanalysisofarealexperimentinwhichweaddmultiplesubstancessimultaneouslytothesensortofindthecontributionofeachsubstanceindependently.

The absolute value of sn can be determined by combining the ratio‐based

approachwith the theoreticalmodel used for the theoretical approach (see section

4.4).Inthisway canbedetermined.However,thisrequiresagaintuningofalotof

2ThisequationisusedforFig5.2.Fortheanalysisoftheotherdatabasedontheratio‐based

approachweusedaslightlydifferentshapeofmatrix1

R

(seeAppendix4.A)whichyieldsslightlydifferentresultsintermsofscalingof sn ,whichmakescomparisonoftheabsolute

valueof sn fromexperimenttoexperimentnotpossible.Theshapeof sn overtimeisthe

same.


parameters. Therefore, itmight be easier to determine scaling parameters for sn

using calibration experiments, which are usually also carried out by a singlewavelength.Alternatively,thismethodcanalsobeusedasaquickscreener,providingonlyayesornoansweronthepresenceoftheanalyteandwhichthereforedoesnot

requireknowingtheabsolutevaluesof sn .

An advantage of the ratio‐based method is that direct discrimination betweendifferent substances is possible, also when sn ’s occur in same layer in the

evanescentfield,aslongastheratiosinwavelengthsaredifferentfromeachotherandas long as the substances do not influence each other’s ratio. The analysis of twosubstanceswhichinducea sn isrelativelyeasyasonlytworatiosaredeterminable

parameters which determine the shape of the signal of sn over time. For three

substancesinducinga sn ,alreadysixratiosneedtobedetermined, illustratingthe

rapidly increasing complexity when increasing the number of distinguishablesubstances inducinga sn .However, these ratioscanbedetermined independently

foreachsubstancetodiscriminate thesesubstances fromeachother.Thiswasdone

for threedifferentwavelengths ( 1 =457nm, 2 =561nmand 3 =660nm),and

threedifferent substances:85nmcarboxylatedpolystyrenebeads,proteinAandD‐glucoseasshowninsection4.3.2.

4.3.2EffectiverefractiveindexratiosmeasuredatvariouswavelengthsTheratios /

m

k m

sR areexperimentallydeterminedforthreedifferentwavelengths( 1 =

457nm, 2 =561nmand 3 =660nm),and for threedifferentsubstances:85nm

carboxylated polystyrene beads, protein A and D‐glucose, which are substancesmeasured inChapter5, representing specific binding, non‐specificbinding andbulkchangesrespectively.Fig.4.1showsanexampleofameasurementwithD‐glucoseand85nmbeadsmeasuredat 1 =457nmand 3 =660nm, illustratingthedifference in

660/457R for both substances. A single value for the ratio for a given substance is

determinedbycalculatingthemeanvalueoftheratiosoverthetimeintervalinwhichthe ratios have a relatively stable value for that substance. In practice itwas found

thatthetimeintervalsthatshowastableratiocorrespondtointervalsinwhich1,effN

wasbetween80‐100%ofitsmaximumvalueforD‐glucose,20‐100%for85nmbeads and 10 ‐ 100 % for protein A. Therefore these intervals were applied fordetermination of all ratios. The determined ratios are presented in Table 4.1. The

82 Chapter4

error margins of the ratios are determined by twice the standard deviation (95%confidenceinterval)ofthemeasuredratiosfromdifferentexperiments.

Moreresearchisrequiredtodeterminewhattheoriginofthespreadoftheratiosis.Inordertodothis,itshouldbeinvestigatedifthespreadcanbeexplainedbytheartefactsorothernoisesourcesofthemeasurements(seeChapter3),thecleaningofthesurfacesof thechip(largerspread isseen inbindingofsubstancescomparedtobulkchangesduetoD‐glucose)orthesubstancesself(spreadinproteinAissmallercompared to beads). For each substancewhich should be distinguished from othersubstances,theratiosshouldbemeasured.

Fig.4.1:The effN measuredat457nm(blueline)andat660nm(redline)whenadding6.16

mg/mlD‐glucoseand1.0μg/ml85nmcarboxylatedpolystyrenebeadstothesensorplottedontheleftaxisandthecorrespondingratio 660/457R (orangeline)plottedontherightaxis.

Table4.1:Experimentallydetermined effN ratiosat 1 =457nm, 2 =561nmand 3 =660nmdeterminedforD‐glucose,85nmcarboxylatedpolystyrenebeadsandproteinA.

660/457R 561/457R 660/561R

D‐glucose 1.218±0.016 1.160±0.012 1.050±0.00885nmbeads 0.918±0.045 0.998±0.037 0.920±0.040ProteinA 0.680±0.022 0.830±0.022 0.820±0.023

4.4Combinedanalysisapproach

Tobeabletodeterminetheabsolutevaluesof sn oracorrespondingsurfacemass

coveragesC the theoretical approachand ratio‐basedapproachwere combined.Forthiscombinedapproach,themeasuredratiosof effN atmultiplewavelengthscanbe

0 500 1000 1500 2000

0.0

0.5

1.0

1.5

2.0

1.0g/ml85nmbeads

PBS

PBS

PBS

Neff

Time(s)

660nm457nm

R660/457

6.16mg/mlD‐glucose

0.50

0.75

1.00

1.25

1.50

R660/457

x10‐4


fitted to the theoretical ratios of effN at multiple wavelengths by changing the

waveguide parameters. Assuming thatD‐glucose causes a homogeneous sn in the

evanescentfieldofthesensor,themeasuredratioscanbefittedtotheexpectedratiosbasedonthetheoreticalmodelbychangingwaveguideparameters(e.g.ncore,nsub,dcore)such that the expected andmeasured ratios correspond.Next, using thewaveguideparametersdeterminedfromtheD‐glucoseratios,theratiosofthesubstanceswhichdo not induce a homogeneous sn can be fitted by varying the layer thicknessd1,

such that we can say that the sn occurred in that layer. Consequently, the

determinedvaluesforwaveguideparametersandd1canbeinsertedinthetheoreticalmodelandcombinedwiththemeasured effN ’s todeterminetheabsolutevaluesof

sn .

Theratioscanbefittedbychangingthencoreandnsub,resultinginanoptimalfitfor0.068coren and 0.04subn with respect to the specifications of RI of the

materials. Although the RI of the materials can be slightly different due tocontaminations, these differences in RI of these materials (SiO2 and Si3N4) are notrealistic.Thetheoreticalratioswerealsofittedwiththemeasuredratiosbychangingnsol.However,nsolcouldnotbechangedsuchthatthemeasuredandtheoreticalvaluescorrespond.Alternatively,itispossibletofitthetheoreticalvalueswiththemeasuredratiosbychangingthewavelengthofthelightsources,butthewavelengthshadtobechangedintheorderoftensofnanometrestogiveafitwithinthestandarddeviationsof themeasured ratios. This is very unrealistic, because the lasers have a specifiedwavelengthwithin 0.3 nm [1, 2]. The ratios can also be fitted by changing thedcorebetween 62.4 nm and 63.4 nm remaining within the standard deviation of themeasuredratiosforD‐glucose.Theoptimalfitwasfoundfordcore=62.9nm,basedon

660/457 660/457 561/457 561/457mean mean mintheory meas theory measR R R R .Thesevaluesarefroma

fabricationperspectivenotveryrealistic,becausethespecifiedthicknessofthedcoreofthe used waveguide is 70 1 nm. Instead of fitting only one parameter, it is alsopossibletofittheratiosbytuningmultipleparameters,whichisplausiblebecauseitisverylikelythatthewaveguidebehaviourisaneffectofthesumofmultipleerrorsinspecified waveguide parameters. However, even then it is not guaranteed that theexact values are found and that this results in a better fit. Therefore, we base ourresults on the best fitwithdcore=62.9 nm (the fit ofncore andnsub resulted both in ahighercomparedtothefitwithdcore)andassumethatourwaveguidebehaveslikeawaveguidewithaneffectivedcore=62.9nm.

84 Chapter4

Fig. 4.2: Measured ratios /

m

k m

sR (dots, diamonds and triangles) for protein A, 85 nm

carboxylatedpolystyrenebeadsandD‐glucoseforallcombinationsofthewavelengths457nm,561nmand660nm.Alsoindicatedistheaverageofthemeasuredratiosandthecorrespondingerror bars representing the 95% confidence interval. The pink squares represent thetheoreticallydeterminedratios forahomogeneous n basedondcore=70nm,andall layersstartingatthesurfacewithathicknessofdglucose=infinity,dprotein=2nmanddbead=85nm.Theblack linesarethetheoreticalratios fittedtothemeasuredratio,bychanging first thedcore to62.9tofittheratiosoftheD‐glucoseandchangingd1to2.1nmand72.6nmtofittheratiosoftheproteinAandthe85nmbeadsrespectively.

Next, the ratios, bead660/457R and bead

561/457R can be fitted to the theoretical model by

changingthethicknessdofthelayerwhichstartsat0(surface)andendsatdandinwhichthe n occurred,assumingahomogeneous n inthislayer.Thisresultedinanoptimalfit( min )fordbead=72.6nm.Fittingsuchthatthetheoreticalratiosstayed

withintheerrormarginsofthemeasuredratiosresultedinaminimumandmaximum

dbead of57.5nmand88.4nmrespectively. In similarway fittingprotein660/457R and protein

561/457R ,

resultedinoptimaldproteinof2.1nmandamaximumdproteinof7.3nm.Thosevaluesareclosetotheexpectvaluesof1nm[3]to2.5nm[4]fortheproteinAand85.0±6.5nmforthebeads.Differencesfortheoptimalfittedbeadthicknessmightbeexplainedbythefactthattheydonotgiveahomogeneous n ,becausetheyaresphericalinsteadof the assumed cubical particles.Moreover, differencesbetween the theoretical andexperimental waveguide configuration and artefacts in the measurements can alsoleadtodifferencesbetweenthedeterminedthicknessandrealthicknessofthebeads.Moreresearchisrequiredtodeterminethecorrectnessofthedeterminedthicknesses.The measured ratios from Table 4.1, the theoretical ratios (based on specifiedwaveguideparameters)andthefittedtheoreticalratiosbasedondcore=62.9nm,dbead=72.6nmanddprotein=2.1nmareshowninFig.4.2.Theindividualpointsplottedin

thesamefigurerepresentthemean /m

k m

sR foreachsinglemeasurement,allperformed

0.6

0.8

1.0

1.2

1.4

R

85nmbeadsProteinA D‐Glucose

660/457561/457

660/561

Theoreticalvalues

Fittedvalues


withthesamewaveguide.Thefittedthicknessesdbead,dproteinanddcorecanbeinsertedin the theoretical approach to determine scaling factors which can be used todeterminetheabsolutevalueof n forthesesubstancesandacorrespondingsurfacemass coverage of the beads and the proteins. Finally, we can say the averagemeasuredratioscorrespond to the theoretical ratioswhenchoosingdcore=62.9nm,dbead=72.6nmanddprotein=2.1nm.Moreover,themeasuredratiosshowaspreadandaswasstatedbefore,moreresearchisrequiredtodeterminetheoriginofthisspread.Thespreadislowenoughsuchthatratiosofthedifferentsubstancesdonotoverlap.Consequently, it is possible to discriminate between the different substances usingtheseratios.Themoredifferenttheratios,thelesssingulartheanalysismatrix,whichmeansthatthedrift,theaforementionedartefactsandnoisewillbelessenhanced.

4.5InfluenceinputparametersinanalysisapproachesondeterminedRIchange

Fig.4.3: Effective refractive index change over time at 1 = 457 nm and 2 = 660 nm for

additionof6.16mg/mlD‐glucoseonlyandadditionof6.16mg/mlD‐glucoseand1.0μg/ml85nmbeadssimultaneously.AfteraddingD‐glucoseandapplyingawashingstepthe effN comes

backtoitsbaselineasexpectedastheD‐glucosedoeshardlybindtothesurfaceandwillfullybereplaced by the PBS buffer again.. After adding beads and D‐glucose simultaneously andapplying gain a washing the step the effN does not go back to zero because beads stay

attachedtothesurfaceofthechip.An extensive study on the application of the analysis approach on differentmeasurementscanbe found inChapter5.Here,weonlypickoutonemeasurementwhichwewillnotdiscussindetailbutonlyusetoillustratetheinfluenceoftheinputparametersofthetheoreticalandratio‐basedapproachesontheendresultof n .Fig.

0 500 1000 1500 2000 2500

0

1

2

3

4

PBS6.16mg/mlD‐glucose

Neff

Time(s)

660nm457nm

x10‐4

PBS

PBS6.16mg/mlD‐glucose+1.0g/ml85nmbeads

86 Chapter4

4.3 shows the measurement where first D‐glucose only and later on 85 nmcarboxylated polystyrene beads plus D‐glucose simultaneously were added tophosphatebufferedsaline(PBS)inthesensor.Forthisanalysisitisonlyimportanttoknowtheexpectedendresultandtheinfluenceoftheinputparametersonthisresult,which will be discussed for the theoretical approach and ratio‐based approach insections4.5.1and4.5.2respectively.Wefocushereontheshapeof n andnot theabsolutevaluesof n .

4.5.1TheoreticalapproachFor the theoretical approach we determine

1ln and

2ln which should increase

equallywhenonlyD‐glucoseisadded(betweent≈500sandt≈1000s)independentofthechosenlayerthicknessd1asD‐glucoseisexpectedtogiveahomogenous n inthewhole evanescent field.When awashing step at t ≈ 1000 s both

1ln and

2ln

shouldgobacktozeroasalltheD‐glucoseisreplacedbyPBSagain.Whenbeadsareaddedthed1isideallychosenequallytotheheightofthebeadssuchthatthe n duetothebeadsonlyoccurs in layer1andthereforecanbediscriminated fromthe n due to theD‐glucose.Consequently,whend1 is chosenright,

2ln shouldgoback to

zeroafterthewashingstepisappliedatt≈2000sandthe1ln shouldbeequaltothe

n duetothebeads.The theoretical analysis approach was applied to the experiment of Fig. 4.3 to

illustratetheinfluenceofthetheparametersdcoreandd1ontheendresultof n .Hereweusethefollowingstrategy.Varyingthedcoreresultsinavaryingleveloftheboththevalue of n of layer 1 and layer 2 at the D‐glucose step. Independent of the layerthickness d1, the n at the addition of only D‐glucose should be the same in bothlayers, which is realized by choosing dcore = 63 nm. Again, these values are from afabricationperspectivenotveryrealistic,becausethespecifiedthicknessofthedcoreofthe usedwaveguide is 70 1 nm.However, otherwaveguide parameterswere alsotunedtofitthemeasurementdata,butdidnotgivebetterresults.Alternatively, it isalsopossible to tunemultiplewaveguideparameters,but also thenguaranteed thatthe exact values of the waveguide parameters are found and that this results in abetterfit.Therefore,wecharacterizeourchipsbyintroducingahomogeneous n duetoadditionofD‐glucoseandfitthisbychangingthedcoreofthewaveguide.


Fig. 4.4: Determined n with the theoretical analysis approach varying a, b, c) the corethicknessdcore andd,e, f) the thicknessof the first layerd1, illustrating the influenceof inputparameterson theendresultof n Layer1(L1)start startsatzeroandendsatd1,whereaslayer2(L2)startsfromd1andendsatinfinity.Theblacklineisthe n oflayer1minusthe n of layer 2 and represents the n due to the binding of the beads assuming a low surfacecoverageofthebeads.

Next,theendlevelofbothsignalscanbetunedbyvaryingd1.Theendlevelofthe

bulkshouldbezero,becausetherewasno n forthebulkcomparedtothestartingsituationwerethebulkwasalsoPBS.Inthecaseoflineardrift,theendlevelwillnotbezerobutinlinewiththelineardrift.Thiswasrealizedwithad1=65nm,assumingadcore=63nm.Next,weinvestigatetheresults ifwechooseslightlydifferentvaluesforthesetwoparameters.Fig.4.4showssomeexamplesofdeterminedsignalsof n

using different values of dcore and d1. It shows that the ln is influenced by the

differentinputparameters.Changingthecorrectlyfitteddcore=63nmwith1nmandusingthesamed1resultsinpercentagewisedifferencein 1 2n n (blackdottedline)

ofmaximally 2%.Withdifferent fittedd1’s this can increase. Therefore, aD‐glucosecharacterizationstepwasappliedineachmeasurementtoverifytheresponseofeachchipandtobeabletodetermineitsdcore.

0 1000 2000

0

1

2

0 1000 2000

0

1

2

0 1000 2000

0

1

2

0 1000 2000

0

1

2

0 1000 2000

0

1

2

0 1000 2000

0

1

2

d e f

cbx10‐3n

Time(s)

L1(0‐74nm)L2(74nm‐inf)L1minusL2

dcore=62nma dcore=63nm

n

Time(s)


x10‐3 x10‐3 dcore=64nm

n

Time(s)


n

Time(s)


x10‐3 dcore=63nm

n

Time(s)


x10‐3 dcore=63nm

n

Time(s)


x10‐3 dcore=63nm

88 Chapter4

4.5.2Ratio‐basedapproachFor the ratio‐based approach the beadn should not changewhen only D‐glucose is

added(betweent≈500sandt≈1000s), beadn shouldincreasewhenthebeadsand

D‐glucoseareaddedsimultaneouslyatt≈1500sand beadn stabilisewhenawashing

stepisappliedatt≈2000s.The glucosen shouldincreasewhenD‐glucosewasadded

at t ≈ 500 s and t ≈ 1500 s and come back to approximately its zero levelwhen awashingstepwasappliedatt≈1000sandt≈2000s.

Fig.4.5: Determined n with the ratio‐based analysis approach varying a) the ratio glucose

/k lR

resultinginachangeofthebeadssignalandb)theratio bead/k l

R resultinginachangeoftheD‐

glucosesignal.

Inthiscasetheratio 660/457beadR wasvariedfrom0.885till0.905andtheratio glucose

660/457R

wasvariedfrom1.210till1.230.TheresultsareshowninFig.4.5.Theratio 660/457beadR

determinesthesignaloftheD‐glucoseandtheratio glucose660/457R determinesthesignalof

thebeads.Inthiscasethebestfitforthesignalofthebeadswasfoundfor glucose660/457R =

1.220,havingazerolevelforthebeadswhenonlyD‐glucosewasadded.Tofittheendlevel of theD‐glucosewith the expectedvalue, assuminga constant lineardrift, the

660/457beadR =0.895.Again,differentvaluesresultindifferentamplitudesofinthiscasethe

sn .Adifferencein glucose660/457R of0.02resultsina beadn ofapproximately4%,whereasA

differenceof0.02in 660/457beadR resultedinanapproximately14%differencein glucosen .

Therefore,itisimportanttoknowhowaccuratelytheratioscanbedetermined.Moreresearch is required to determine what is causing the spread in the ratios as wasstatedbefore.

0 1000 2000

0

2

4

6

8

10

0 1000 2000

0

2

4

6

8

10x10‐5 b

n u

nscaled

Time(s)

D‐glucose85nmbeads,Rglucose

660/457=1.21

85nmbeads,Rglucose660/457

=1.22

85nmbeads,Rglucose660/457

=1.23

a x10‐5

n u

nscaled

Time(s)

85nmbeadsD‐glucose,Rbead

660/457=0.885

D‐glucose,Rbead660/457

=0.895

D‐glucose,Rbead660/457

=0.905


4.6Influenceofnoise,driftandartefactsondeterminedRIchange

Simulationsweredonetodeterminetheinfluenceofdrift,noiseandartefacts(duetoboundaryeffectsofshiftinginterferencepatterns,seeChapter3)inameasuredsignalof effN on theresultof n .For thedeterminationof n , thematrixof theratio‐

basedanalysisapproach isusedtogetherwith 660/457bindingR =0.919 , 660/457

bulkR =1.217.Fig.

4.6aandbshowasignalof effN andthecorresponding n whennodrift,noiseand

artefacts are added to the signal. Next, Poisson distributed noise with differentstandard deviations (see labels in Fig. 4.6) is added to the signal. Fig. 4.6c/d showconfirmstheresultsfromChapter2thatnoisein effN isenhancedintheresult n .

Artificial fluctuations (similar as artefacts due to boundary effects of shiftinginterferencepattern, seeChapter3)werealsoadded to thesignalof effN andalso

showupenhancedintheresult n astheyarenotseenin effN butareseenin n ,

especiallyinthesignalforthebindingasshowninFig.4.6e/f.Theinfluenceoflineardriftin effN on n isshowninFig.4.7.Lineardriftinthe

signal effN is translatedtoa lineardrift intheresultof n .Thesignof the linear

driftof n isthedeterminedbythesignandratioofthedriftof effN atthedifferent

wavelengths.Thiscanbeexplainedfromatheoreticalperspective:

,1 1 ,11 1

,2 2 ,22 2

eff eff

eff eff

N c t Nn cM M M t

N c t Nn c

, (4.7)

where c1 and c2 are constantswhich determine the amplitude of the linear drift of

effN .Onecanseethatthisfinallyresultsinlineardriftinthesignalof n whichis

determinedbythecoefficientsofthematrixMandconstantsc1andc2.

90 Chapter4

Fig.4.6: a) Simulated signal of effN over time comparablewith a typicalmeasurement but

without anynoise andb) the corresponding unscaledn over timedeterminedwith ratio‐based

approach,c)thesimulatedsignalof effN overtimewithnoiseofdifferentstandarddeviations

(seeblackoutlinedboxesforzoomin)andd)thecorresponding unscaledn overtimeillustrating

theenhancementofnoiseinthesignalof n ,e)the effN overtimewithaddedartefacts(black

line)andf)thecorresponding unscaledn overtime.

0 200 400 600 800 1000

0

1

2

3

4

0 200 400 600 800 1000

0

2

4

6

0 200 400 600 800 1000

0

1

2

3

4

5

6

0 200 400 600 800 1000

0

2

4

6

8

0 200 400 600 800 1000

0

1

2

3

4

0 200 400 600 800 1000

0

2

4

6

d

fe

c

Nodrift,noiseandfluctuations

Neff

Time(s)

457nm660nm

x10‐4

x10‐5 Nodrift,noiseandfluctuations

n u

nscaled

Time(s)

BindingBulk

x10‐4

Neff

Time(s)

457nm,noise2x10‐6

660nm,noise2x10‐6

457nm,noise2x10‐7

660nm,noise2x10‐7

457nm,noise2x10‐8

660nm,noise2x10‐8

x10‐5

Binding,noise2x10‐6

Binding,noise2x10‐7

Binding,noise2x10‐8n u

nscaled

Time(s)

Bulk,noise2x10‐6

Bulk,noise2x10‐7

Bulk,noise2x10‐8

x10‐4

bNeff

Time(s)

457nm660nmAddedartefacts

a

x10‐5

n u

nscaled

Time(s)

BindingBulk


Fig. 4.7: a, c, e, g) A effN over time with various added linear drifts and b, d, f, h) the

correspondingdetermined unscaledn forbinding andbulk signal, determinedwith ratio‐based

approach,illustratingthatlineardriftin effN resultsisenhancedlineardriftin unscaledn .

0 200 400 600 800 1000

0

1

2

3

4

5

0 200 400 600 800 1000

0

2

4

6

8

0 200 400 600 800 1000

0

1

2

3

4

0 200 400 600 800 1000‐2

0

2

4

6

8

0 200 400 600 800 1000

0

1

2

3

4

0 200 400 600 800 1000

0

2

4

6

8

0 200 400 600 800 1000

0

1

2

3

4

0 200 400 600 800 1000‐8

‐4

0

4

8

12

x10‐4N

eff

Time(s)

457,drift+2x10‐8s‐1

660,drift+2x10‐8s‐1

457,drift‐2x10‐8s‐1

660,drift‐2x10‐8s‐1

x10‐5

nunscaled

Time(s)

Binding,drift457nm,660nm:2x10‐8s‐1

Bulk,drift457nm,660nm:2x10‐8s‐1

Binding,drift457nm,660nm:‐2x10‐8s‐1

Bulk,drift457nm,660nm:‐2x10‐8s‐1

x10‐4

Neff

Time(s)

457,drift2x10‐8s‐1

660,drift1x10‐8s‐1

x10‐5

nunscaled

Time(s)

BindingBulk

x10‐4e

g h

f

d

b

c

Neff

Time(s)

457,drift1x10‐8s‐1

660,drift2x10‐8s‐1

a

x10‐5

nunscaled

Time(s)

BindingBulk

x10‐4

Neff

Time(s)

457,drift1x10‐8s‐1

660,drift‐2x10‐8s‐1

x10‐5

nunscaled

Time(s)

BindingBulk

92 Chapter4

4.7Discussionandconclusions

In this chapterwe showed two analysis approacheswhich also could be combinedintoathirdanalysisapproach.Thetheoreticalapproachisexact,butrequirestuningofmany parameters to determine the correct values for the sensitivity coefficientsthatarerequiredtodeterminethecorrectvalueof n andsurfacemasscoverageofananalyte.Hence,thismethodisexactbutinpracticedifficulttoimplementbecauseofthemanyparametersthathavetobetuned.Therefore,wedevelopedamuchmorepracticalratio‐basedapproachbasedontheratiosofthemeasured effN ’smeasured

at different wavelengths. These ratios can be determined independently for eachsubstancetodiscriminatethesesubstancesfromeachother.Thiswasdoneforthreedifferentwavelengths( 1 =457nm, 2 =561nmand 3 =660nm),andthreedifferent

substances:85nmcarboxylatedpolystyrenebeads,proteinAandD‐glucosesuchthatit canbeused for theapplicationof thisapproach in thenext chapter.On theotherhand,usingthisratio‐basedapproachitisnotpossibletodetermineanabsolutevalueof n as scaling factors are unknown. However, to determine e.g. an analyteconcentration it is not required to know these, because calibration measurements,whichareusuallyalsocarriedoutbyasinglewavelengthYI,canbeusedtocalibratetheunscaledvaluesof n with an analyte concentration.Additionally, thismethodcan also be used as a quick screener, providing only a yes or no answer on thepresenceof theanalyte.Todetermine theabsolutevalueof n ora correspondingsurface mass coverage, a combined theoretical and ratio‐based approach wasdeveloped,whichwasalsousedtodeterminethicknessof85.0±6.5nmbeadstobeinbetween 57.4 nm and 88.4 nm andwith an optimal fit of 72.6 nm. Protein Awasdetermined to be smaller than 7.3 nm, with an optimal fit of 2.1 nm which is inrelativelygoodagreementwiththemeasuredthicknessesof1nm[3]and2.5nm[4].

Toapplythetheoreticalapproachorcombinedapproachamoreelaboratestudyis required to characterize thewaveguide structure as itwas shown that the inputparametershaveagreatinfluenceontheoutcomeof n orthicknesses.Welldefinedbeads(insizeandRI)withdifferentsizescanbeusedtostudyifthereisastructuraldifferencebetweenthetheoreticalandexperimentallydeterminedsizes.Additionally,alternativesforD‐glucosewhicharewelldefinedinRIandgiveahomogeneous n ,and do not bind the sensor surface, can be used to determine the waveguideparameters. These characterization experiments should also be used to test therobustnessoftheanalysismethods.

Finally,theinfluenceofnoise,driftandartefactsin effN ontheresultof n was

tested. The simulations confirmed the calculations in Chapter 2which showed thatthere isa trade‐offbetweensize‐selectivityandsensitivity.Simulationsshowedthatnoise(e.g.shotnoise)andartefacts(oscillationswhichoccurduetoboundaryeffects


ofshiftinginterferencepattern)in effN arefoundbackenhancedintheendresultof

n ,whichmeansthatartefactsandnoiseshouldbeaslowaspossible.Furthermore,itwas shown that simulated linear drift in effN results in enhanced linear drift in

n .Soassuminglineardriftin effN ,itispossibletocompensatedforthatin n .In

this casewe used the ratio‐based approach to test the influence of noise, drift andartefactsin effN ontheendresultof n .However,theinfluencewillbesimilarfor

thetheoreticalapproachasbothapproachesuseasimilarmethods(bothuseamatrixmultiplication)todetermine n from effN .

Acknowledgements


Appendix4.ASlightlydifferentmatrixforratio‐basedapproach

Here,we show the slightlydifferent formsof thematrix1

R

whichareused for theanalysisofsomeexperimentsusingtheratio‐basedanalysisapproach.Theexpressionofequation(4.6)ofthemaintextwasfoundinthefinalstageofthisprojectsuchthat

notimewaslefttoapplythematrixofthisequationtoallthemeasurements.The1

R

whichwasusedatanearlierstageinthisprojectcanalsobeinsertedinequation(4.6)

ofthemaintextandwillbegiveninthisappendix.Theslightlydifferentmatrices1

R

resultinslightlydifferentresultsintermsofscalingof sn ,whichmakescomparison

of the absolute value of sn from experiment to experiment are used not possible

when differentmatrices are used. However, the shape of the signals of sn are the

same.Themaintextstatesforwhichfigures1

R

ofthemaintextisused.Theanalysisof other experiments analysed with the ratio‐based approach are based on thematriceswhicharegiveninthisappendix.

Forthesituationoftwosubstanceswhichshouldbediscriminatedbasedon effN

measuredattwowavelengths,theslightlydifferentmatrixisgivenby:

94 Chapter4

2

2 1

1

2 1

1 /

/

11

11

s

s

RR

R

. (4.A.1)

For thesituationof threesubstanceswhichshouldbediscriminatedbasedon effN

measuredatthreewavelengths,theslightlydifferentmatrixisgivenby:

3 32 2 2

2 1 2 1 3 1 2 1 2 1

2 3 3 3 32 2 2 2

2 1 3 1 2 1 2 1 3 1 3 1 2 1 2 1 3 1

3 1 1

2 1 2 1 3 1

1 3 1 1

2 1 3 1 2 1 3 1

/ / / / /

/ / / / / / / / /

1 / / /

/ / / /

111 1

11

s ss s s

s s s s ss s s s

s s s

s s s s

R R R R R

R R R R R R R R R

R R RR

R R R R

3 1

2 1 2 1

3 3 31 1

2 1 3 1 2 1 3 1 2 1

2 1 1 2 1

2 1 2 1 3 1 2 1 2 1

1 2 1 1 2 2 1 1

2 1 3 1 2 1 3 1 2 1 3 1 2 1 3 1

/ /

/ / / / /

/ / / / /

/ / / / / / / /

11

111 1

s s

s s ss s

s s p s s

s s s s s s s s

R R

R R R R R

R R Rs R R

R R R R R R R R R

2

2 1/s

. (4.A.2)

References

1. CoboltAB,“OwnersManualModel05‐01”(CoboltAB,February2015,2015),retrieved01‐06‐2015,http://www.cobolt.se/wp‐content/uploads/2015/03/Owners‐Manual‐05‐01_150206‐1.22.pdf.

2. CoboltAB,“OwnersManualModel04‐01”(CoboltAB,March2015,2015),retrieved01‐06‐2015,http://www.cobolt.se/wp‐content/uploads/2014/10/Owners‐Manual‐04‐01_v1.63_20150305.pdf.

3. M.C.Coen,R.Lehmann,P.Groning,M.Bielmann,C.Galli,andL.Schlapbach,“AdsorptionandbioactivityofproteinAonsiliconsurfacesstudiedbyAFMandXPS,”J.ColloidInterfaceSci.233,180‐189(2001).

4. S.Ohnishi,M.Murata,andM.Hato,“CorrelationbetweenSurfaceMorphologyandSurfaceForcesofProteinAAdsorbedonMica,”Biophys.J.74,455‐465(1998).

Chapter51

Experimentalapplicationofanalysisapproachesforsize‐selectiveanalytedetectionAbstractIn this chapterwepresent theexperimentalapplicationof thesize‐selectiveanalyteapproaches presented in Chapter 4. We use the different analysis approaches todiscriminatebetweenbindingof85nmbeadsfrombindingofproteinA,andbindingof85nmbeadsorproteinAfromD‐glucosebulkchangesbymeasuring effN attwo

wavelengths.Threewavelengthswereusedtodiscriminatebetweenthreesubstancessimultaneously, however this led to noisy results which leaves the applicabilityuselessatthismoment.Forasuccessfulapplicationoftheseanalysisapproachesusedto determine three independent n ’s, the artefacts (due to boundary effects of theshiftinginterferencepatternorlensorgratingaberrationsintheimagingpathofthesetup)anddrift in effN shouldbereducedsignificantly.Alternatively, itwasshown

thatothertechniquestoimprovespecificity,suchastheuseofareferencechanneltocompensate for bulk changes, can be used next to size‐selective detection. Theworking of the ratio‐based approach to discriminate between two substances wasverified by a blind experiment with samples containing different concentrations ofprotein A and 85 nm beads. Furthermore, a triplicate ofmeasurements serieswithdifferentbeadconcentrationsandaconstantproteinAconcentrationshowedthatwecoulddiscriminaterefractiveindexchangescausedbybindingofproteinAand85nmbeads,evenwhenthe n oftheproteinwasapproximately20timeshigher.Thiscanforexamplebeusedtodiscriminatespecificanalytebindingof largerparticles fromnon‐specificbindingofsmallerparticlestoimprovetheperformanceoftheYIsensorandIOinterferometricsensorsingeneral.

Partofthischapterisusedinmanuscript:H.K.P.Mulder,C.Blum,V.Subramaniam,J.S.Kanger,“Size‐selectiveanalytedetectionwithaYounginterferometersensorusingmultiplewavelengths”,manuscriptinsubmission

96 Chapter5

5.1Introduction

In this chapterwe first test the reliability of the sensor in section 5.2 (testedwithsetup with end‐fire coupling, see Chapter 3). Next, we apply the different analysisapproachestomeasurementswith85nmcarboxylatedpolystyrenebeadswhichbindto the sensor surface (representing specific binding of e.g. viruses which haveapproximately this size), protein A which also binds to the sensor surface but issmaller(representingnon‐specificbinding)andD‐glucosewhichhardlybindstothesensorsurfacebut inducesabulkchange.Thesemeasurementsweredonewith thefiberbutt‐endcouplingaspresented inChapter3.These substances induce n ’s atdifferentplaceswithintheevanescent fieldandcanthereforebediscriminatedfromeachother.Allthemeasurementsaredoneusinga1xphosphatebufferedsaline(PBS)buffer towhich the different sampleswere added. Themeasurements startswith abaseline of approximately 400 s to determine the drift of effN signal. This can be

usedtocompensatefordrift,aslineardriftin effN resultsinlineardriftin n (see

Chapter4).Allthegraphsof effN and n overtimewhichareshowninthisthesis

arerawdata( effN )ordirectlydeterminedfromtherawdata( n )withtheanalysis

approaches,soarenotcorrectedfordrift.Onlythedeterminedvaluesof effN and n

atonetimepointinthegrapharecorrectedforlineardriftbyfittingalinearcurvetothebaselineof400swhich is shown in thegraphsandsubtracting thevalueat thetimepointofinterestbasedonthelineardriftfromthevalueoftherawdataatthatsametimepoint.Tominimizetheeffectonlineardriftandartefacts(duetoboundaryeffectsofmovinginterferencepattern),theappliedconcentrationsarerelativelyhighcompared to these drift and artefacts. Next, the measurements continue with acharacterizationstep,containinganadditionof6.16mg/mlD‐glucosefor500stothePBS,whichcanbeusedtoseeifthesensorrespondsasexpectedortofitatheoreticalmodel by changing waveguide parameters such that the determined curves of n fulfil the expectations (see Chapter 4). In section 5.3we discriminate between twodifferent substancesusing the theoretical and ratio‐basedapproach, afterwhichwediscriminatebetween threesubstancesusing thesameapproaches insection5.4. Insection5.5weshowthatothertechniquestoimprovespecificityofthesensorcanbeusednexttothesize‐selectivedetectionusingmultiplewavelengths.Next, insection5.6weverifytheratio‐basedapproachbyaddingunknownconcentrationsof85nmbeadsandproteinAtothesensorinablindexperimentanddiscriminatingthemfromeach other. Subsequently, in section 5.7 the combined approach is applied todiscriminatebetween85nmbeadsandproteinAanddetermineitsabsolutevaluesofn andthecorrespondingsurfacemasscoverage.Finally,weendthischapterwitha

discussionandconclusions.

Experimentalapplicationofanalysisapproachesforsize‐selectiveanalytedetection 97

5.2Reliabilityofsensor

To determine the reliability of the sensor with the simultaneous detection of zeroorder TEmodes ofmultiple wavelengths, nine concentration series with D‐glucoseweremeasured.FortheconcentrationseriesD‐glucosesolutionsof6.16mg/ml,12.32mg/mland18.48mg/mlconcentrationweresuccessivelyaddedtoMilli‐Qwater.Thiswas done in three different channels and the experimentwas repeated three timesusingadifferentchipeachtime.AtypicalexampleofonemeasurementinonechannelisshowninFig.5.1a.

Fig.5.1:a)Ameasurementof effN overtimeforaconcentrationseriesofD‐glucosemeasured

at spatial frequency d14 in measurement M3, b) the corresponding effN for all the three

measurements comparedwith the theoretical effN . For visibility reasons, thedatapointsof

the three different measurements are plotted slightly offset, but they all correspond to theglucoseconcentrationsof6.16,12.32or18.48mg/ml.

Next, the effN duetoD‐glucose isdeterminedbythedifferencebetween effN

whenD‐glucosewas added to the chip and thevalueof effN which itwouldhave

basedon lineardriftwhennoD‐glucosewouldhavebeen added. Subsequently, the

effN is plotted in Fig. 5.1b as a function of the glucose weight percentage.

Furthermore, the theoretical effN [1] is plotted as a function of the weight

percentage inFig.5.1b.Forvisibilityreasons, thedatapointsbelongingto the threedifferentmeasurements(M1,M2,M3)doneondifferentchipsareslightlyoffset,buttheyallcorrespondtotheglucoseconcentrationsof6.16,12.32or18.48mg/ml.Equalsymbolsbelong tomeasurementsdoneon the samechip,butmeasuredatdifferent

0 2000 4000 6000‐1

0

1

2

3

4

5

6

7

5 10 15 201

2

3

4

5

6

7

x10‐4

Neff

Time(s)

redd14greend14blued14




H2O

H2O

H2O

H2O

b660nm,M1561nm,M1457nm,M1660nm,M2561nm,M2457nm,M2660nm,M3561nm,M3457nm,M3660nm,theory561nm,theory457nm,theory

x10‐4

Neff

D‐glucoseconcentration(mg/ml)

a

98 Chapter5

channels. Themeasured effN corresponds relativelywell to the theoretical values

for effN .However,thereisaspreadinthemeasuredvaluesfor effN whichcanbe

caused by slightly different core thicknesses or sensing window lengths of thedifferent chips. Chips were fabricated from the same wafer. The spread betweenchannelsissmallerthanthespreadbetweenchips.Furthermore,smalldifferencesin

effN may be attributed to artefacts due to boundary effects of the shifting

interferencepattern(seeChapter3).Itcanbeconcludedthatthesemeasurementsshowthatthesensorincombination

withthemultiplewavelengthworksasexpected,butthatcomparativemeasurementsshould preferably be done with a single waveguide and preferably in the same

channel, because of the different response of effN in each different channel and

waveguide.Alternatively,measurements canbedoneondifferent chipsor channelsby applying a characterization step where D‐glucose only is added to the sensorbeforemeasuringtherealsample(seeChapter4).

5.3Discriminationbetweentwodifferent‐sizedsubstances

In this sectionwepresentmeasurements inwhichwediscriminatebetween twoofthe three named different‐sized substances. First, 6.16 mg/ml D‐glucose and 2.0

μg/mlproteinAwereaddedsimultaneouslytothesensor(seeFig.5.2a)and effN is

measuredat 1 =457nmand 2 =660nm.Asacharacterizationstep6.16mg/mlD‐

glucoseonlywasaddedtothesensor,whichresultsinanincreaseofthe effN asD‐

glucosehasahighern thanPBS.Asexpected, the effN is thehighestat 2 as it is

moresensitivetobulkchanges(D‐glucosedoeshardlybindtothesurface)compared

to 1 .Afterapplyingawashingstep, the effN comesbacktozeroastheD‐glucose

bulk solution is completely replaced by PBS. After simultaneously addingD‐glucose

andproteinA the effN increases againdifferently at 1 compared to 2 . For both

effN ’s the increase consist of a steep slopedue to thebulk effect anda less steep

slopeduetothebindingoftheproteinAtothesurface.AfteragainapplyingawashingstepthesignaldropsduetothefactthattheD‐glucoseandproteinAinthebulkare

replacedbyPBS.The effN doesnotgobacktozeroasproteinAisleftatthe

surface.Inthiscasethe effN at 1 ishigherthanat 2 whichisexpectedsince


Fig.5.2:The effN measuredovertimebyaddingtothesensorsurface:a)85nmcarboxylated

polystyrene beads and D‐glucose, d) protein A and D‐glucose and g) protein A and 85 nmcarboxylated polystyrene beads and corresponding determined n using the theoreticalanalysisapproach(b,e,h)andtheratio‐basedanalysisapproach(c,f,i).shorterwavelengths aremore confined to the core of thewaveguide and thereforemoresensitive tochangescloseto thesurfacecomparedto longerwavelengths.Fig.5.2dshowsasimilarexperimentwhereafteradditionofonly6.16mg/mlD‐glucose,6.16mg/ml D‐glucosewas added simultaneouslywith 1.0 μg/ml 85 nm beads. Fig5.2gshowsasimilarmeasurementwhereafteradditionofonly6.16mg/mlD‐glucose,1.0μg/mlproteinAwas added simultaneouslywith2.0μg/ml85nmbeads. In thenext subsections, the n ’s of these different measurements determined with the

0 500 1000150020002500

0.0

0.5

1.0

1.5

2.0

2.5

0 500 1000150020002500

0

2

4

6

0 500 1000150020002500

0

5

10

15

20

0 500 1000150020002500

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0 500 1000150020002500

0.0

0.4

0.8

1.2

1.6

0 500 1000150020002500

0

5

10

15

20

0 500 1000150020002500

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0 500 1000150020002500

0

4

8

12

0 500 1000150020002500‐20

‐15

‐10

‐5

0

5

10

Ratio‐basedapproachTheoreticalapproach

i

f

c

e

b

hg

d

Neff

Time(s)

660nm457nm

x10‐4


PBS

PBS

PBS6.16mg/mlD‐glucose+2.0mg/mlproteinA

Measurementx10‐3

nTime(s)

Layer1(0‐3.5nm)Layer2(3.5nm‐infinity)Layer1minuslayer2

x10‐5

nunscaled

Time(s)

ProteinAD‐glucose

6.16mg/mlD‐glucose+1.0g/ml85nmbeads


x10‐4

Neff

Time(s)

660nm457nm

PBS

PBS

PBS

x10‐3

n

Time(s)

Layer1(0‐70nm)Layer2(70nm‐infinity)Layer1minuslayer2

a

x10‐5

nunscaled

Time(s)


x10‐4

Neff

Time(s)

660nm457nm


2.0g/ml85nmbeads+1.0g/mlproteinA

PBS

PBS

PBS

x10‐3

n

Time(s)

Layer1(0‐3.5nm)Layer2(3.5nm‐infinity)

x10‐5

nunscaled

Time(s)

proteinA85nmbeads

100 Chapter5

theoretical and ratio‐based approach. The signal determined from the ratio‐basedapproach is called unscaledn as this signal is related to a n induced by the

aforementionedsubstances.However,informationtheabsolutevalueislostusingtheratio‐based approach as ratio of a substance is constant independent of itsconcentration(assuminglowconcentrationsandthereforenoaggregationofproteinandbeads on topof eachother). Therefore, only the shapeof unscaledRI change ( unscaledn )andnottheabsolutevaluesof unscaledn ’sshouldbecomparedwiththe n

’softhetheoreticalmodel.

ProteinAandD‐glucoseFig. 5.2b shows the n corresponding to Fig. 5.2a determinedwith the theoreticalapproach. Assuming a homogeneous n in the whole evanescent field for the D‐

glucose characterization step, the determined1ln should be the same as

2ln ,

independent of the arbitrarily chosen thickness of the layers. The dcore, whichdeterminesthetheoreticalsensitivitycoefficientsandtherefore n ,canbeadjustedsuch that both n ’s of both layers for the glucose step reach an equal levelindependentoftheselectedlayers(seeChapter4).Thebestfitof n wasachievedbyadcoreof61.8nm.The n forthefirstD‐glucosestepbasedonthisdcoreof61.8nmis

equalto 48.78 10 RIUwhichagreeswellwiththe 48.88 10 RIUdeterminedfromthe

effN measuredat660nmassumingthesamedcore.

Next to thedcore, also the layer thicknessof the first layer (d1),ofwhichonecan

determinea n ,canbeadjusted.Changingd1resultsinavaryingendlevelof1ln and

2ln as layer 1 is defined from 0 ‐ d1 and layer 2 from d1 – infinity. In the case of

simultaneousadditionofD‐glucoseandproteinA,theexpectedendlevelof2ln after

thewashingstepbacktoPBSiszero.Thelayerthicknessthatprovidestheexpectedbulk levels for theexperimentwith theproteinAandD‐glucose is3.5nm,which isreasonably close to the determined height of an absorbed layer of protein A ofapproximately1nm[2]‐2.5nm[3].Assuminglowconcentrationsandbulkchanges

occurringinbothlayers,the2ln wassubtractedfrom

1ln resultinginasignalwhich

shouldbelongtoonlytheproteinA(seeblackdottedlineinFig.5.2b).TheproteinAsignalattheendisapproximatelyequaltoaseventeenthofafullcoverageofproteinA.

The determined n corresponding to Fig. 5.2a and based on the ratio‐based

approachisshowninFig.5.2c.Theratios protein660/457R and glucose

660/457R thatwereusedforthis

ratio‐basedapproacharetakenfromTable4.1.Amplitudesof smn canbecompared

with smn of other measurements, however sm

n should not be compared with


amplitudes of snn as scaling is required (see Chapter 4). The glucosen fulfils the

expectationsasthesignalgoesupwhenD‐glucosewasaddedandcomesdowntothebaseline when applying a washing step. The proteinn goes up when protein A was

added and slowly decreases after applying a washing step, which is probably becaused by partly desorption of the protein A from the sensor surface. However,

proteinn goesalsoslightlyupwhenD‐glucosewasadded.Withan glucose660/457R of1.220the

proteinn goesbacktotheexpectedlevelbasedonlineardriftof proteinn .This glucose660/457R of

1.220remainswithin theerrormarginsgiven inTable4.1whicharea resultof thedifferencesinthemeasuredratiosofthesamesinglesubstancesmeasuredatdifferentexperiments.

BeadsandD‐glucoseInasimilarwayas for theproteinAandD‐glucose,Fig.5.2dwasanalysedwiththetheoreticalmodel as shown inFig.5.2e.Theoptimal resultwas realizedbasedonadcoreof63.4nmandd1of70nm,whichislowerthantheexpected85±6.5nm.Thismightpartlybeexplainedbythefactthatthebeadsaresphericalandnotcubical,astheresponseof the effN ’s inducedbyacubicshapedparticlebindingonthesensor

surface issimilartotheresponseofthe effN ’s inducedbya largersphericalshaped

bead binding on the same sensor surface. By subtracting the2ln from

1ln and

assuminglowconcentrationsandbulkchangesoccurringinbothlayers itwasagainpossibletodetermineasignalwhichshouldbelongtoonlythebeads.

Usingtheratio‐basedapproach, glucosen and beadn weredetermined,againbased

onthemeasuredratiosofTable4.1.Asexpected,the beadn doesnot increasewhen

adding D‐glucose, increases when adding the beads and remains constant afterapplyingawashingstep.The glucosen reachesalevelwhichisapproximatelythesame

asinFig.5.2casexpectedbecausetheaddedD‐glucoseconcentrationswerethesame.However,afterthewashingstepthe glucosen isabitlowerthanexpected.Anoptimal

resultisgivenat bead660/457R =0.900whichagainremainswithinmeasurederrorbarsof

bead660/457R .

ProteinAandbeadsForthemeasurementofsimultaneousadditionofbeadsandproteinAthetheoretical

approachisnotverysuitable,especiallybecause2ln isfarfromhomogeneousasthe

evanescent field penetrates deeper than the thickness of the bead. Moreover, theexpectedendlevelsoftheproteinAandthebeadsarenotknownwhichwasknownforthebulkeffectoftheD‐glucose.Consequently,thefittingofd1isnotfeasible.

102 Chapter5

Usingtheratio‐basedapproachitispossibletodiscriminatebetweentheproteinAandthebeads.TheresultbasedonratiosofTable4.1isshowninFig.5.2i.Aswasmentioned, amplitudes between proteinn and beadsn should not be compared, as

scaling factors of substances are individual. Calibration measurements or thecombined approach are required for that. However, proteinn of Fig. 5.2i can be

compared with proteinn of Fig. 5.2c. In Fig. 5.2i half of protein A was added to the

sensor, but the measured proteinn is only 1.6 times lower. An even lower signal of

proteinn isexpectedasnowalsobeadscancoverpartofthesurfacewherebeadscan

bind.Furthermore,the beadsn ofFig.5.2fisapproximately2.6timeshighercompared

to beadsn ofFig.5.2i,wheretheaddedbeadsconcentrationwastwotimeslower.The

differencesmightbeexplainedthatdifferentchipsanddifferentstocksolutionswereusedforthedifferentmeasurements.Furthermore,allthemeasurementsarebasedonthe non‐specific adhesion of the protein A and the beads to the sensor surface.Therefore,smalldifferencesbetweensensorsurfacescanmakealargedifference.Theresult from D‐glucose in which no binding is involved, the signals were in niceagreementwitheachother.

Besides,Fig.5.2ishowsthatusingtworatios,excluding glucose660/457R ,itisnotpossible

to suppress the first D‐glucose step from the protein and beads signal. Threewavelengths are required to discriminate between protein A, 85 nm beads and D‐glucoseinasinglemeasurement,aspresentedinthenextsection.

5.4Discriminationbetweenthreedifferent‐sizedsubstances

To test if both analysis approaches are also applicable when using threewavelengths to discriminate three n ’s induced by three different substances, D‐glucose (6.16 mg/ml), 85 nm carboxylated polystyrene beads and protein A wereaddedtothesensoraftereachother.The effN wasmeasuredat 1 =457nm, 2 =561

nmand 3 =660nmasshowninFig.5.3a.Fig.5.3bshowsthe n correspondingto

Fig. 5.3a determined with the theoretical approach after fitting the dcore, d1 and d2basedonanalysingtheresultsbasedontwowavelengths.AnexampleofthisisshowninFig.5.3cwheredcore andd2weredeterminedas62.7nmand64nmrespectively,basedon 1 and 3 . It can be seen that at theD‐glucose step the 1l

n and2ln are

equal and when adding the beads at t3 there is an increase of1ln and

2ln stays

approximately constant. Similarly, the fittingwas donewith 1 and 2 , and for the


combinationofD‐glucoseandproteinAandthecombinationofproteinAandbeads,basedonthewavelengthcombination 1 with 2 and 1 with 3 .This resulted ina

dcore=62.55±0.15nm,d1=5±2nmandd2=60±4nm, illustrating thatdifferentwavelengthcombinationsgivedifferentfitvalues,meaningthattheparametersinthetheoreticalmodelareprobablynotcompletelyright.TheresultinFig.3bisbasedontheaveragevaluesfordcore,d1andd2andthreewavelengthstodiscriminatebetweenthe n ofthreelayers.Theoverallsignalsdeterminedwiththetheoreticalapproachbased on three wavelengths/layers are much more noisy compared to the twowavelengths/layerscase,whichmakestheoptimizationmoredifficult.Theincreaseinnoise,artefactsanddriftcanbeexplainedbythefactthatthematrixtodetermine

1ln

from effN becomes more singular and therefore all noise sources are enhanced

which was already shown in Chapter 2. However, ignoring the noise, drift andartefacts,allsignalsincreaseroughlywiththesameamplitudewhenaddingD‐glucoseat t1 (

1ln has a much higher amplitude, so is plotted on a different scale). When

addingthebeads,anincreasein1ln and

2ln isexpected,butthereisalsoanincrease

of3ln .AlsowhenaddingproteinAatt5 thereisanunexpecteddecreaseof

2ln and

unexpected increaseof3ln as itwasonlyexpected that

1ln would increase.These

observations probablymean that the signals1ln ,

2ln and

3ln are not sufficiently

decoupled.Optimizingthetheoreticalmatrixfurtherismorecomplexcomparedtothetwo wavelengths/layers situation, because changing one parameter will result inchangeofall thesignals

1ln ,

2ln and

3ln , illustrating thecomplexityand limitsof

the theoretical approach.Therefore,we ratio‐basedapproach isbetterused for thisanalysis.

Fig. 5.3d shows the determined sn with the ratio‐based approach. The ratios

used for the analysis are again based on the determined ratios for the individualsubstances.Betweent1andt2only glucosen changesapartfromsomefluctuationsseen

for beadn and proteinn , which can be explained by boundary effects of the shifting

interference pattern (see Chapter 3). At t3 the beadswhere added, but besides theincreaseof beadn thereisalsoanincreaseof proteinn and glucosen whichmeansthat

thesignalsof sn arenotperfectlydecoupled.Thisisalsothecaseatt5whenprotein

A isaddedtothesensorwhichmainlyresult inan increaseof proteinn ,butalso ina

smallincreaseof glucosen .Therefore,theratioswerealsooptimizedbyanalysingthe

measurementbasedontwowavelengths/substances.AnexampleisshowninFig.5.3f

104 Chapter5

Fig.5.3: a)The effN measuredover timestartingwithaPBSbuffer, adding6.16mg/mlD‐

glucoseatt1resultinginhigher effN forthelongerwavelengthsasexpected.Afterapplyinga

washing step at t2, 1.0 μg/ml 85 nm beads was added at t3. The response of the shorterwavelengthsisnowlarger.AfterwashingagainwithPBSatt4,2.0μg/mlproteinAwasaddedatt5 resulting in a relatively stronger response of the shorter wavelengths. After applying awashingstepatt6thesignaldecreasesduetodesorptionoftheproteinA,b)thecorresponding

ln determinedwith the theoretical approach fittingdcore,d1 andd2 based on analysing the

measurementwithtwowavelengths(seeexampleshowninc)),the sn determinedwithratio‐

based approachbasedond) the ratiosmeasuredwith the individual substances and e) aftertuning the ratios based on analysing the measurement with two wavelengths and twosubstancesofwhichanexampleisgiveninf).

0 1000 2000 3000

0

1

2

3

4

5

0 1000 2000 3000

0

1

2

3

0 1000 2000 3000‐2

0

2

0 1000 2000 3000‐1

0

1

2

3

4

0 1000 2000 3000‐1

0

1

2

3

0 1000 2000 3000

0

4

8

12

16

20

t3

t2

t1

Neff

Time(s)

660nm561nm457nm

x10‐4

t4

t5

t6

x10‐4x10‐2

c

b

n layer1

Time(s)

Layer1(0‐3nm)

t4

t1

t3

t2

t6

t5

a

d

e f

0

2

4

6

8

10

L2(5‐60nm)L3(60nm‐inf)

n layer2,3

n

Time(s)

Layer1(0‐64nm)Layer2(64nm‐inf)

t2t

1

t3

t4

t6

t5

x10‐3 x10‐5

n u

nscaled

Time(s)

ProteinA85nmbeadsD‐glucose

t2

t1

t3

t4

t6

t5

x10‐5

n u

nscaled

Time(s)


t2

t1 t

3

t4

t6

t5

x10‐5

n u

nscaled

Time(s)

ProteinAD‐glucose

t2

t1

t3

t4

t6

t5


whichshows that proteinn wassufficientlydecoupled from glucosen basedon 1 and

3 . Because only twowavelengthswere used to discriminate between proteinn and

glucosen itwasnotpossibletodiscriminatethemfrom beadn whichisshownbetween

t3 and t4,where both proteinn and glucosen increase. Fig 3e shows the result of sn

after optimization of the ratios which led to protein561/457R =0.837, protein

660/457R =0.698, bead561/457R

=0.967, bead660/457R =0.885, glucose

561/457R =1.158and glucose660/457R =1.215 fromwhich s

660/561mR canbe

calculatedbydividing s660/457mR by s

561/457mR .Allthesevaluesfallwithintheerrormargins

of the experimentally determined ratios of the individual substancespresented in Table 4.1. Using the slightly changed ratios, the three substances arenicelydecoupled,illustratingthatitispossibletodiscriminatebetweenthreedifferent

substances. However, the enhanced drift and artefacts in sn makes it difficult to

analysetheseresults.Moreover,whensubstanceswereaddedsimultaneously,fittingbasedontwosubstances/twowavelengthsisnotpossible.Therefore,theratiosusedfortheanalysisshouldbemeasurablefrommeasurementswiththesinglesubstancesandnotbe tunedafterwards.Torealise this thespread inratiosshouldbereduced.Moreresearchisrequiredtodeterminewhattheoriginofthespreadoftheratiosis.In order to do this, it should be investigated if the spread can be explained by theartefactsorothernoisesourcesofthemeasurements,thecleaningofthesurfacesofthechip(largerspreadisseeninbindingofsubstancescomparedtobulkchangesdueto D‐glucose) or the substances self (spread in protein A is smaller compared tobeads).

TheproteinA,85nmbeadsandD‐glucosewerealsoaddedsimultaneouslytothesensorwhichisshowninFig5.4.InthefirstmeasurementproteinAwasaddedtothemeasurement channel and in the secondmeasurement protein Awas added to thereference channel such that there is no interference between binding of protein Aandbindingofbeads.Fig.5.4a, c shows thedetermined effN measuredat457,561

and660nm.TheD‐glucosestepwasusedtodeterminethe glucose660/457R ,becauseanother

chipwasusedforthismeasurement.Next,themeasuredratiosofthesubstancesfromTable4.1areusedandcorrectedpercentagewisebasedonthedifferenceinratiosofD‐glucose measured on this chip and in Table 4.1 (multiplied by

glucose,currentchip glucose,Table4.1660/457 660/457R R ).

Fig 5.4b, d show the n determined with ratio‐based approach, based on thepercentagewisecorrectedratios.Scalingfactorsarechosensuchthat proteinn , beadn

and glucosen havesimilaramplitudesandcanbeseenclearlyinthegraph,butshould

notbecompared.Againitisclearthatthe sn determinedforthreedifferent

106 Chapter5

Fig. 5.4: a) The effN measured over time where 6.16 mg/ml D‐glucose was added

simultaneouslywith1μg/ml85nmbeadsand1μg/mlproteinAtoPBS,b)the effN overtime

forameasurementwhere6.16mg/mlD‐glucosewasaddedsimultaneouslywith1μg/ml85nmbeadsto1xPBSinthemeasurementchanneland1μg/mlproteinAto1xPBSinthereferencechannelandb,d) thecorresponding n forproteinA,beadsandD‐glucosedeterminedwiththejointtheoreticalandratio‐basedapproach.Thedashedlinesarefittedtothebaselinesofthesignal assuming linear drift and neglecting the artefacts due boundary effects of the shiftinginterferencepattern.

substancesgivesnoisysignals.TheaforementionedartefactsthatshowupinthedataareintheorderofthesignalofthebeadsandproteinA.Neglectingtheartefacts,thedashedlinesinthegraphsof n arefittedtotheassumedlinearphasedrift.Overall

the signals of sn correspond roughly to the expected signals, however the drift

artefactsareinthesameorderofmagnitudeasthesignal.Therefore,itisnotpossibletoanalyse beadn and proteinn properly.Thus,withthecurrentsetupandcurrentdrift

0 500 1000 1500 2000 2500‐1

0

1

2

3

0 500 1000 1500 2000 2500‐0.2

0.0

0.2

0.4

0.6

0 500 1000 1500 2000 2500‐1

0

1

2

3

0 500 1000 1500 2000 2500‐0.6

‐0.4

‐0.2

0.0

0.2

0.4

0.6

c d

bNeff

Time(s)

660nm561nm457nm

PBS

PBS

6.16mg/mlD‐glucose+1.0g/ml85nmbeads+1.0g/mlproteinA


PBS

x10‐4a

n(a.u.)

Time(s)


Neff

Time(s)

660nm561nm457nm

PBS


PBS

x10‐4

PBS

6.16mg/mlD‐glucose+1.0g/ml85nmbeads+1.0g/mlproteinAinreferencechannel

n(a.u.)

Time(s)



and artefacts it is not possible to discriminate between three n ’s based on threewavelengths.Artefactsanddriftin effN shouldbereducedsignificantlyforsuccessful

application of the size‐selective detectionwhen discriminating between three n ’sbased on three wavelengths. Sufficiently wide optical components behind thewaveguide to image interference patterns on camera (e.g. lenses and grating) arerequired. Alternatively, other techniques to improve specificity can be used next tosize‐selective detection, such that it is only required to discriminate between twosubstances based on twowavelengths. For example, to discriminate the binding ofbeadsfrombindingofproteinAandbulkchangesduetoD‐glucose,D‐glucosecanalsobeaddedtothereferencechanneltocanceloutthecontributionduetoD‐glucoseasisshown in the section 5.5. In that case only two wavelengths are required todiscriminatebetweenbindingofbeadsandproteins,resultinginlessenhancementofnoise.

5.5Simultaneoususeofothertechniquesnexttosize‐selectivedetection

In this sectionwe illustrate thepossibility to use alternative techniquesnext to thesize‐selective detection in order to discriminate between e.g. two instead of threesubstances. For example non‐specific binding can be reduced by using specificantibodiesandblockingagents.Moreover,abulkcontributioncanbecancelledoutbyapplyingthesamebulksolutioninmeasurementandreferencechannels.Anexampleofsimultaneousadditionofabulkcontributiontoameasurementchannel(channel2)andreferencechannel(channel4)isshowninFig.5.5.Fig.5.5ashowsthe effN

of457,561and660nmovertimebelongingtospatialfrequencycorrespondingtod12(channelcombination1‐2)whenadding2.0μg/ml85nmbeads,0.5μg/mlproteinAand 6.16 mg/ml D‐glucose simultaneously to channel 2. Fig. 5.5b shows thecorresponding n contributionof thebeads, protein andD‐glucoseusing the ratio‐basedapproach.Thecharacterizationstep(notshowninFig.5.5)resultedinafitfordcore = 68 nm. Consequently, the ratios used for the analysis were again the ratiosmeasured in Chapter 4 (see Table 4.1) and percentagewise corrected based on the

measureddifferenceinratiofortheD‐glucosestep.Fig.5.5bshowstheresulting sn ,

whichisaverynoisesignal,mainlycausedbytheaforementionedartefacts.Fig5.5cshowsthe effN of457,561and660nmovertimecorrespondingtospatialfrequency

d24 (channel combination 2 ‐ 4) when adding 2.0 μg/ml 85 nm beads, 0.5 μg/mlproteinAand6.16mg/mlD‐glucosesimultaneouslytochannel2andsimultaneously6.16mg/mlD‐glucosetochannel4.Thismeasurementwasalsoanalysedwiththe

108 Chapter5

Fig.5.5: a) Themeasured effN over time at 457, 561 and 660nmof the spatial frequency

corresponding tod12adding2.0μg/ml85nmbeads,0.5μg/mlproteinAand6.16mg/mlD‐glucose simultaneously to channel 2, b) the corresponding n ’s of protein A, beads and D‐glucosedeterminedwiththejointtheoreticalandratiobasedapproach,c)the effN overtime

at457,561and660nmofthespatialfrequencycorrespondingtod24adding2.0μg/ml85nmbeads, 0.5 μg/ml protein A and 6.16mg/mlD‐glucose simultaneously to channel 2 and 6.16mg/mlD‐glucosesimultaneously tochannel4,d) thecorresponding n ’sofproteinA,beadsandD‐glucosedeterminedwith the ratiobasedapproachbasedon3wavelengths ande) thecorresponding n ’sofproteinA,beadsdeterminedwiththeratiobasedapproachbasedon2wavelengths.

0 400 800 12000

2

4

6

8

10

0 400 800 1200

0

1

2

3

4

5

0 400 800 1200‐10

‐8

‐6

‐4

‐2

0

0 400 800 1200‐5

‐4

‐3

‐2

‐1

0

0 400 800 1200‐5

‐4

‐3

‐2

‐1

0

PBS

PBS

e

d

c

bd12x10‐4Neff

Time(s)

660nm561nm457nm

2.0g/ml85nmbeads+0.5g/mlproteinA+6.16mg/mlD‐glucoseinCh.2

a x10‐2

n b

eads,D‐glucose

Time(s)


x10‐3

0

1

2

3

4

5ProteinA

n p

roteinA

PBS

2.0g/ml85nmbeads+0.5g/mlproteinAinCh.2,6.16mg/mlD‐glucoseinCh.2andCh.4

x10‐4

Neff

Time(s)

660nm561nm457nm

d24

PBS

n b

eads,D‐glucose

Time(s)

85nmbeadsD‐glucose ‐5

‐4

‐3

‐2

‐1

0

ProteinA

n p

roteinA

x10‐3 x10‐2

n b

eads,D‐glucose

Time(s)

85nmbeads

x10‐3 x10‐2

‐5

‐4

‐3

‐2

‐1

0

ProteinA

n p

roteinA


ratioapproachusingthesameratiosasforFig.5.5bandusingthreewavelengthstodiscriminate between three n ’s (Fig. 5.5d) and using two wavelengths todiscriminatebetweentwo n ’s(Fig.5.5e).Fig.5.5dshowsthatthedeterminedbulkcontribution is approximately zero as expected because the D‐glucose in referenceand measurement channel cancel out each other. Fig. 5.5e shows that when onlyanalysing the signal with two wavelengths to determine the bead and protein A

contribution, the noise in sn is reduced significantly.Moreover, it illustrates that

othertechniquesusedforimprovingspecificityofthesensorcanbeusednexttothesize‐selectivedetectionapproach.

5.6Blindexperimentwith85nmbeadsandproteinA

Toapprovetheusabilityandverifythecorrectnessoftheratio‐basedapproachablindexperimentwascarriedout.SixcombinedsamplesofproteinA(intherangeof0.1‐1.0μg/ml)and85nmbeads(intherangeof0.4‐4.0μg/ml)weremadeandlabelledby an independent person. Beforemeasuring, the samples were carefullymixed toensure homogeneous solutions. Next, the samples were measured in twomeasurement sets performed on the same chip using threemeasurement channelsandone reference channel. In between themeasurement sets thewaveguideswerecleaned using a cleaning protocol presented in Appendix 5.A. A D‐glucosecharacterization step was applied to verify the correct response of the differentwavelengths. This resulted in a different ratio the measured effN ’s of 1.263

comparedtothe1.218±0.016.Thisdifferencecanbeexplainedbythefactthatthesemeasurementswere performedwith a different chipwhich can have for example aslightlydifferentdcore.Tocorrectforthisdeviation,wechangedtheaveragemeasuredratiosofTable4.1percentagewise(multipliedby 1.263/1.218 )andusedthisforthe

ratio‐based approach to analyse this experiment. Fig. 5.6a shows an example of ameasurementwiththeD‐glucosecharacterizationstepandaddingsampleAandFib.5.6b shows the corresponding determined beadn and proteinn over time,which are

readoutbeforeatt=2400swhenthebeadssignalwasconstant.Theexpectedvaluesof n basedonlineardrift(baselinedeterminedbetweent=1320sandt=1380s)issubtractedfromthis 2400stn andplottedasafunctionoftheappliedconcentrationin

Fig.5.6candFig.5.6dforthe85nmbeadsandtheproteinArespectively.The applied bead concentration and the determined beadn show a linear trend

which means the beadn was determined correctly using the ratio‐based approach,

assumingalinearrelationbetweenthenumberofbeadsinthesampleandresponseofthesensor(seeAppendix5.Bforverification).Moreover,itshowsthatitispossible

110 Chapter5

Fig. 5.6: Blind experiment with six samples (A‐F) containing 85 nm beads and protein Ashowinga)themeasured effN overtimeofonemeasurementofthesetofblindexperiments

b)thecorrespondingdetermined beadn and proteinn usingtheratio‐basedapproachandc)the

determined n asafunctionofthebyforehandunknownappliedconcentrationsofthe85nmbeads and d) the determined n as a function of the by forehand unknown appliedconcentrationsof theproteinA. Forvisibility reasons, thedatapointsA andFof the appliedproteinconcentrationareslightlyoffsetbuttheybothcorrespondto0.9μg/ml.

todoacalibrationexperimenttofindtherelationsbetweenthebeadsconcentrationandtheresponseofthesensor.Furthermore,alsoalineartrendisseenbetweenthemeasured proteinn and the applied protein concentration. Only at higher applied

proteinAconcentrations, thedetermined beadn flattensoff,whichcanbeexplained

by the fact that for thesemeasurements the surface was saturated with protein AwhichcanbeseeninFig.4b.Thatthesignalof beadn stillincreasesmightbe

explainedbythebeadsreplacingtheproteinAofthesurface.Thebeabletocompare

0 1000 2000

0

2

4

6

8

10

0 1000 2000‐6

‐4

‐2

0

2

4

6

8

0.0 0.2 0.4 0.6 0.8 1.00

2

4

6

8

0 1 2 3 40

1

2

3

Neff

Time(s)

660nm457nm

x10‐4

PBS

6.16mg/mlD‐glucose sampleA

PBS

PBS

n p

rotein(a.u.)

Time(s)

proteinA

c d

‐1

0

1

2

x10‐4 x10‐5

85nmbeads

n b

ead(a.u.)

x10‐4x10‐5

E

A

FD

B

C

Measuredn(a.u.)

Appliedconcentration(g/ml)

ProteinA

b

85nmbeads

B

E

D

F

C

Measuredn(a.u.)


A

a


Fig. 5.7: Blind experiment with six samples (A‐F) containing 85 nm beads and protein A,showingthedeterminedsurfacemassconcentrationsasafunctionofthebyforehandunknownappliedconcentrationsofa)the85nmbeadsandb)theproteinA.

beadn and proteinn , the signals were both read out at the same time point. The

amplitude of the signals of beadn , proteinn and the corresponding surface mass

coveragecanbecomparedbyapplying thecombinedapproach.Theabsolutevalues

beadn and proteinn were determined using the optimal fits of dcore, dbead and dprotein

foundinChapter4.ThecorrespondingCbeadwasdeterminedassumingcubicalbeadswith an average determined height of 72.6 nm, a density of 1.05 g/cm3 [4] and

561nm 1.595polystyrenen [5], 589nm 1.334PBSn [6]andusingequation2.6

fromChapter2assumingthattheRIchangeisgivenby beadn insteadof 2 3n n as

was determined for the layer case. The corresponding Cprotein to proteinn was

determinedassumingproteinAwithaneffectivesizeof2.1nm(=averagedeterminedheight),amassof42kDa[7]and 1.41proteinn [8].TheCproteinandCbeadareplotted in

Fig.5.7whichshowsthatitwaspossibletodiscriminatethebindingofbeadsfromthebinding of proteins, evenwhen theCprotein was approximately 20 times higher thanCbead(sampleA).

5.7Validationandreproducibilityofsize‐selectivedetection

Tovalidate theworkingof the size‐selectivedetectionmethod and todetermine itsreproducibilityatriplicateofameasurementserieswhere85nmbeadsandproteinAwereaddedsimultaneouslytothesensorwasperformed.Samplesweremadewithavarying bead concentration (0, 0.75, 1.5 and 3 μg/ml) and a fixed protein Aconcentration of 0.5 μg/ml and added to the sensor. The combined approach was

0.0 0.2 0.4 0.6 0.8 1.0

0

2

4

6

8

0 1 2 3 4

0.0

0.4

0.8

1.2

1.6

A

FD

B

E

C

C(ng/mm

2 )


ProteinAb85nmbeads

B

E

D

F

C

C(ng/mm

2 )


A

a

112 Chapter5

applied to with determine how much beads we can discriminate from how muchprotein A in terms of absolute RI changes and corresponding surface coverages.Insertingtheoptimalfitsofdcore,dbeadanddproteinfoundinChapter4inthetheoretical

modelandusing protein660/457R =0.680and beads

660/457R =0.918resultedinadetermined beadn

and proteinn shown Fig. 5.8. Fig. 5.8a shows a typical example of the combined

approachappliedonameasurementwhereproteinAand85nmbeadswereaddedsimultaneously.The beadn and proteinn weredeterminedatt=1050swhenthesignal

of beadn was stable. The signal of theproteinAwas not always stable at this time

point. Sometimes the protein A was still increasing (see Fig. 5.8a) and sometimes,whennobeadswereaddednexttotheproteinA,alreadysaturated,sothe beadn of

the different measurements cannot exactly be compared. However, to be able tocompareCproteinandCbead, thesignalof beadn and proteinn werereadoutatthesame

time.Furthermore,bothwerecorrectedforalineardriftbasedonthefirst400s.Thedetermined beadn and proteinn forallmeasurementsofthetriplicateareshowninFig.

5.8b. Fig. 5.8b shows that each measurement series has a linear trend (R2≥0.996)whichwasexpectedasitwasshownthatatwotimeshigherbeadconcentrationleadto approximately twice asmanybeadsbound to the surface (seeAppendixB).Thisverifies that we can use size‐selective detection to discriminate binding of 85 nmbeads sufficiently frombindingofproteinA.The signalmeasured for theprotein isdecreasingwithanincreasingbeadconcentration.Thismightbeexplainedbythefactthat for higher bead concentrations more beads are covering the sensor surface,resultingislessspacefortheproteinstobind.

Despite the threedifferencesshowa linear trend, there isadifference(which issmallerthanafactortwo)betweenthetriplicates.ThemeasureddifferencebetweenthetriplicatesmightbeexplainedbythefactthatfewerbeadsandproteinAbindtothesurfaceofthechipifitwasnotcleanedsufficientlybetweenruns.Thesamechiphadtobereused,andthuscleaned(seeAppendix5.Aforcleaningprotocol)betweenexperiments,tobeabletodetermineadcoreanddswhichcanbeusedfortheanalysisofall the experiments. As was mentioned before, different chips resulted in differentratios, possibly due to slightly different core thicknesses or RI’s or as a result ofartefacts in themeasurements. To be able to use different chips for reproducibilitystudiesartefactsshouldbereducedandifstilldifferencesaremeasuredwithdifferentchips, improvement of the chips is required. Carboxylated fluorescents beads wereusedtoverifythatthebeadsweresufficientlyremovedaftercleaning(seeAppendix5A).However,itisnottrivialtocheckforanyresidualproteinAasthesurfaceofthesensingwindowisembeddedinthewaveguideandtheusedproteinAissmallerthanthediffraction limitandnot fluorescent, and thusnoteasy tovisualise.However, astheproteinAisdesorbingmorecomparedtothebeadswhenapplyingawashingstep,itisexpectedthattheproteinAisremovedsufficientlybythisprotocol.Nevertheless,


it should be noted thatwe use non‐specific adhesion of particleswhichmeans thatsmall surfacechangescan lead todifferentamountsofbindingofproteinAand thebeads.Alternatively,itispossiblethatbeadsorproteinAsticktothetubewheretheyarestoredbeforethemeasurementstartsortothetubeswhichbringthesubstancestothesensorsurface.Thelatterargumentcanalsoresultinabaselinewhichslightlyincreasesduetothefactthatanalyteswhichwereattachedtothetubewallsduringpreviousexperimentsbecomelooseandsubsequentlyaretransportedtothesensingwindow. To test the influences of these factors, measurements should be repeatedwheresubstancesarekeptintubesanequalamountoftimebeforemeasurementandtubesshouldbereplacedforeachmeasurement.Inthesemeasurementsthiswasnotthecaseandonlysubstancesintubesweremixedbypipettingthesolutioninandoutof the tubes just before adding to the sensor. Moremeasurements are required todeterminetheoriginofthespreadofthetriplicates.

Fig.5.8:Exampleof n determinedwithcombinedapproachforameasurementwhere85nmbeadsandproteinAwhereaddedsimultaneouslyatthesensorandb)thedetermined n forthreemeasurementserieswhereproteinAand85nmbeadswhereaddedsimultaneouslywithavaryingbeadconcentrationandafixedproteinconcentrationof0.5μg/ml.

TodeterminemaximaldistinguisheddifferenceinsurfacemasscoverageCinthis

measurement (which is not equal to themaximal distinguishable C in general), wedeterminehere theC (in a similarwayasbefore) corresponding tohighest ratioof

protein beadsn n andusingformula2.D.5(seeAppendix2.D),resultinginCbeads= 23.3 10

pg/mm2andCprotein =5.6ng/mm2.Thismeans85nmbeadsarediscriminated fromproteinAwith an approximately 17 timeshigherC.Wedidnotpush it to the limithere,soinprincipleevenahigherdifferencebetweensurfacemasscoveragecouldbereached as was already shown in the blind experiment. The maximal resolvabledifference inCwhich is different for eachmodel system, so for eachmodel systemmeasurements are required to determine the maximal resolvable difference in C.

0 500 1000 1500‐0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

0.0 0.5 1.0 1.5 2.0 2.5 3.0

0

2

4

6x10‐2

PBS

n

Time(s)

ProteinA85nmbeads

0.5g/mlproteinA+1.5g/ml85nmbeads

PBS

b

R2=0.998

R2=1.000

x10‐2

Measuredn b

eads

Appliedbeadconcentration(g/ml)

M1,beadM2,beadM3,beadLinearfit

R2=0.996

a

0

2

4

6M1,proteinM2,proteinM3,protein

Measuredn p

rotein

x10‐3

114 Chapter5

Furthermore,itwasnotpossibletoverifythenumberofbeadsandproteinboundtothechipsurfaceastheywerenon‐fluorescentandsosmalltheyarenoteasilyimaged.


Bymeasuringthe effN attwodifferentwavelengths itwaspossibletodiscriminate

betweentwodifferentsubstancesinducinga n intheevanescentfieldofthesensor.Thebindingof85nmbeadsorproteinAwasdiscriminated fromD‐glucoseandthebinding of 85 nm beads was discriminated from the binding of protein A. Thedetermined n induced by D‐glucose corresponds to the n due to D‐glucose asmeasuredwithasinglewavelengthwhereasthe n ’sduetoboundproteinAand85nmbeadstothesurfacecouldnotbeverified.TheproteinAandbeadscouldnotbeimagedastheyaresmallerthanthedetectionlimitandnon‐fluorescent.However, itwasnotourgoal todeterminetheexactamountofboundproteinAorbeadstothesurface. Using three effN ’s to discriminate between three different substances

inducinga n ,theresultbecamemorenoisybecausethedriftandartefactsin effN

(duetoboundaryeffectsoftheshiftinginterferencepattern)showupmoreenhancedwhen discriminating between three n ’s based on effN ’s measured at three

wavelengthscomparedtotwo n ’sbasedon effN ’smeasuredattwowavelengths,as

the analysismatrixwhich is used for this becomesmore singular. For a successfulapplicationof theseanalysisapproachesusedtodeterminethree independent n ’s,the artefacts anddrift in effN shouldbe reduced significantly.Alternatively, other

techniquesto improvespecificitycanbeusednexttosize‐selectivedetection. Itwasshown that when applying equal bulk changes due to D‐glucose in both themeasurementandreferencechannelsitwaspossibletouseonlytwowavelengthstodiscriminatebetweenbindingofbeadsandproteins,resultinginlessenhancementofdriftandartefactscomparedtothethreewavelength/threesubstancescase.

Application of a theoretical analysis approach made it possible to discriminatebetween n ’s occurring in different layers in the evanescent field, assuming ahomogeneous n ineachlayer.However,todeterminethecorrectvaluefor n ,thismethodrequiresmultipleparameters(e.g. thewaveguidecore thickness,waveguideRIs,layerthicknesses,wavelengths)tobetunedthatisdifficultandlabour‐intensive.Therefore, also a much more pragmatic ratio‐based analysis was applied todiscriminate between 85 nm beads, D‐glucose and protein. For the ratio‐basedapproach itwaspossible to usemeasured ratiosof individualmeasurementsof thesubstances (see Chapter 4) to discriminate between the substanceswhenusing thesamechip.UsingadifferentchipiswaspossibletousethemeasuredratioofChapter4whencorrectingthempercentagewisebasedontheratiosin effN measuredwhen


D‐glucose was added only. This was verified by a blind experiment with differentsamples containing different concentrations of protein A and 85 nm beads and atriplicateofmeasurements serieswithdifferentbeadconcentrationsandaconstant

proteinAconcentration. Informationon theabsoluteamplitudeof sn is lostusing

the ratio‐based approach, however for determining an analyte concentrationcalibrationmeasurementscanbeperformedwhichareusuallyalsorequiredusinganYIwithasinglewavelength.Alternatively,theratio‐basedapproachcanbecombinedwiththetheoreticalapproach.Thiswasdonefortheblindexperimentwhichshowedthatwecoulddiscriminated beadn from proteinn , evenwhen theproteinRI change

wasapproximately20timeshigher,resultinginanexpectedlineartrendofthe beadn

withtheappliedconcentration.Theblindexperimentillustratesthatthesize‐selectivedetection using the ratio‐based approach or combined approach can be used todiscriminatespecificanalytebindingof largerparticles fromnon‐specificbindingofsmallerparticles.

Moreover,wecheckedthereproducibilityofthesize‐selectivedetectionusingthecombinedapproachwiththetriplicateof themeasurementserieswith85nmbeadsand protein A. The triplicates showed differences in beadn and proteinn up to

approximatelyafactoroftwo.Thisspreadisnotveryhighifwetakeintoaccountthatour signal is determined by non‐specific adhesion of beads and protein A, whichmeansthatsmallchangesonthesurfacecanaffecttheamountofboundsubstances.Alternatively,itispossiblethatbeadsorproteinAsticktothetubeweretheyarekeptin before the measurement starts or the tubes which bring the substances to thesensor surface. More measurements are required to determine the origin of thespreadofthetriplicatesandtoverifythatitisnotcausedbysize‐selectivedetection.

In summary, we believe that with adding size‐selectivity using the differentpresentedapproaches,wecanstronglyimprovetheperformanceofourYIsensorandIO interferometricsensorsingeneral.Especiallytheratio‐basedapproach isaneasyapproach to distinguish between different substances. With the current setup it ispossibletodiscriminatebetweenmaximallytwosubstancescausinga n ,butwhenthe artefacts in the measurements are reduced significantly it is also possible todiscriminate between three substances and keep a sensitivity comparable to otherbiosensors.

Acknowledgements

IwanttothankXiOPhotonicsB.V.andinparticularRonaldDekkerforthehelpwiththe realization of the fiber butt‐end coupling and Arshdeep Sidhu for her help inpreparing the samples of the blind experiments. This work is supported byNanoNextNL, a micro and nanotechnology consortium of the Government of theNetherlandsand130partners.

116 Chapter5

Appendix5.ACleaningprotocol

Thefollowingstepsdescribethecleaningprocedureofthewaveguidechipsafterandbeforeameasurementwithbeads,proteinAandorD‐glucoseinPBS:

1) rinsechipwithwater2) rinsechipwithacetoneandwashchipwithacetoneandcotton(4x)3) rinsechipwithisopropanol(99.8%,filteredwith0.22μmfilter)andwash

chipwithsameisopropanolandcotton(4x)4) rinsechipwithisopropanol(99.8%,filteredwith0.22μmfilter)5) blowdrywithinstrumentalairornitrogengas(N2)

Fluorescent carboxylated beads were added to the sensor surface and after themeasurementandafterthecleaningprotocolimagedwithafluorescencemicroscopeto verify that the cleaning protocol was sufficient for removing the beads (see Fig.5.A.1).AstheproteinAwasnot fluorescentand issmallerthanthediffraction limit,theproteinAcouldnotbeimagedandthecleaningprotocolwasnotverifiedfortheproteinA.However,asweseethatbeadsbindstrongertothesurfacethanproteinA(for protein A seemore desorption after applying awashing step compared to thebeads),we expect that the proteinA is sufficiently removed from the surface usingthiscleaningprotocol.

Fig.5.A.1: Images of sensingwindows aftermeasurementswith fluorescent beads a) beforeandb)aftercleaning.Bothimagesaretakenwitha10xPlanApoobjectiveandanAndorDU‐885cameraatanexposuretimeof2secondsandamultiplierof80.

Appendix5.BRelationbetweenappliedbeadconcentrationandmeasuredsurfacemasscoverage

Toverifyifthereisalinearrelationbetweenthesamplebeadconcentrationand n and therefore also C, 60 nm carboxylated fluorescent yellow [9] polystyrene beads(SPHEROTMCFP‐00552‐2)wereaddedto thesensor.Aftermeasuring thebindingofthebeadstothesensorsurface(onlyat660nmasitwasshownthatthe457nmlightwasabsorbedby thebeads) theywere imagedwitha fluorescencemicroscope.The


fluidcuvettewasremovedfromthetopofthechipwhichwasturnedupsidedowntobeabletomeasuretheminEPImode.Theyellowfluorescentbeadswereexcitedwitha457nm laser lineand imagedusinga filter cubewith450‐460nmexcitationand470‐500nmemission.

Themeasured (660nm)effN wascalculatedintoasurfacemasscoverage.Alayer

of 60 nm changing from 1x PBS to a homogeneous polystyrene layer

660nm 1.586polystyrenen results in a (660nm)effN . Assuming a tight packing of

beads( 2bead beadV r ),8.49x106beadscanbindtothesurfaceandasinglebeadresults

ina (660nm)effN =4.42x10‐9.Assumingaloosepacking(surfaceareabead= 24 beadr ),

4.45x106canbindtothesurface,resulting (660nm)effN =2.31x10‐9forasinglebead.

The C can now be determined by dividing the measured (660nm)effN by the

(660nm)effN forasinglebeadanddividing thisbythesurfaceareaof thesensitive

partofthesensingwindow(4mmx4μm).Fig. 5.B.1a, c, e show the measured (660nm)effN for three different

concentrationsofbeads.The (660nm)effN wasdeterminedaftersignalwasrelatively

constant(t=1100s)compensatingforlinearphasedriftbasedontheofthefirst300s.Fig.5.B.1b,d, f showanexampleofcorresponding200x200pixels images takenwith the fluorescentmicroscopewith the40Xobjectiveplus1.5magnification.Fourdifferent regions of the sensing window were imaged of which an average Cbead isdeterminedusingalocalmaximaalgorithm(seeendofAppendix),whichiscomparedCbead measured with the YI (see Fig. 5.B.2). The determined surface concentrationscorrespond relatively well. However, it should be taken into account that the fourdifferentareasareonlyasmallpartofthewholesensingregionwhichwasshownnotto be homogeneously covered with beads. Moreover, a spread in intensities wasmeasured for detected beads with the fluorescent microscope. This could not beexplainedbyaspread in fluorescence intensitiesorsizesof thebeadsregardingthecompany. Therefore, it is possible that the detected spots are not single beads butaggregates of beads. The histograms of the intensities did however not show cleardifferentpopulationsofaggregates,whichmightindicatethattherewerenoaggregatesof2,3or4beads,butitmightbepossiblethattherewereaggregatesofahigher number of beadswith a large spread. Thismeans that the exact number ofbeadspersurfaceareacannotbedetermined.However,assumingthattheaggregationisalinearprocesswithrespecttotheconcentrationofthebeads,itcanbeconcludedthat the surface concentration has a linear relationship with the sample beadconcentration. To be sure about the exact surface concentration, dynamic lightscattering, fluorescent correlations spectroscopy or flow cytometry can be used tocheckthesizedistributionofthebeads.

118 Chapter5

Fig.5.B.1:Fluorescentcarboxylatedbeads(60nm)wereaddedindifferentconcentrationstothe sensor showing a, c, e) the measured effN at 660 nm and b, d, f) one of the four

corresponding imagesofapartof thesensingwindow imagedwitha fluorescentmicroscopewitheachredcircleadetectedbeadusingalocalmaximaalgorithm.

0 500 1000 1500‐0.1

0.0

0.1

0.2

0.3

0.4

0 500 1000 1500

0.0

0.2

0.4

0.6

0.8

1.0

0 500 1000 1500

0.0

0.5

1.0

1.5

2.0

2.5

fe

c

PBS

PBS

Neff

Time(s)

0.05g/ml60nmbeads

x10‐5

d

bNeff

Time(s)

PBS

0.10g/ml60nmbeads

PBS

x10‐5

aNeff

Time(s)

PBS

0.25g/ml60nmbeads

PBS

x10‐5


Fig.5.B.2:Thesurfaceconcentrationdeterminedwithafluorescentmicroscopeandcalculatedbasedonameasuredphasechangeat660nmwiththeYI,assumingatightandaloosepacking,allasafunctionoftheinitialbeadsampleconcentration.Localmaximaalgorithmfunction out=pkfnd(im,th,sz) % finds local maxima in an image to pixel level accuracy. % this provides a rough guess of particle % centers to be used by cntrd.m. Inspired by the lmx subroutine of Grier % and Crocker's feature.pro % INPUTS: % im: image to process, particle should be bright spots on dark background with little noise % ofen an bandpass filtered brightfield image (fbps.m, fflt.m or bpass.m) or a nice % fluorescent image % th: the minimum brightness of a pixel that might be local maxima. % (NOTE: Make it big and the code runs faster % but you might miss some particles. Make it small and you'll get % everything and it'll be slow.) % sz: if your data's noisy, (e.g. a single particle has multiple local % maxima), then set this optional keyword to a value slightly larger than the diameter of your blob. if % multiple peaks are found withing a radius of sz/2 then the code will keep % only the brightest. Also gets rid of all peaks within sz of boundary %OUTPUT: a N x 2 array containing, [row,column] coordinates of local maxima % out(:,1) are the x-coordinates of the maxima % out(:,2) are the y-coordinates of the maxima %CREATED: Eric R. Dufresne, Yale University, Feb 4 2005 %MODIFIED: ERD, 5/2005, got rid of ind2rc.m to reduce overhead on tip by % Dan Blair; added sz keyword % ERD, 6/2005: modified to work with one and zero peaks, removed automatic % normalization of image % ERD, 6/2005: due to popular demand, altered output to give x and y % instead of row and column % ERD, 8/24/2005: pkfnd now exits politely if there's nothing above % threshold instead of crashing rudely % ERD, 6/14/2006: now exits politely if no maxima found % ERD, 10/5/2006: fixed bug that threw away particles with maxima % consisting of more than two adjacent points %find all the pixels above threshold %im=im./max(max(im)); ind=find(im > th); [nr,nc]=size(im); tst=zeros(nr,nc); n=length(ind); if n==0 out=[]; display('nothing above threshold'); return; end mx=[]; %convert index from find to row and column rc=[mod(ind,nr),floor(ind/nr)+1]; for i=1:n r=rc(i,1);c=rc(i,2); %check each pixel above threshold to see if it's brighter than it's neighbors % THERE'S GOT TO BE A FASTER WAY OF DOING THIS. I'M CHECKING SOME MULTIPLE TIMES, % BUT THIS DOESN'T SEEM THAT SLOW COMPARED TO THE OTHER ROUTINES, ANYWAY. if r>1 & r<nr & c>1 & c<nc if im(r,c)>=im(r-1,c-1) & im(r,c)>=im(r,c-1) & im(r,c)>=im(r+1,c-1) & ... im(r,c)>=im(r-1,c) & im(r,c)>=im(r+1,c) & ... im(r,c)>=im(r-1,c+1) & im(r,c)>=im(r,c+1) & im(r,c)>=im(r+1,c+1) mx=[mx,[r,c]']; %tst(ind(i))=im(ind(i)); end

0.05 0.10 0.15 0.20 0.25

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35DeterminedusingYI,

densepackingDeterminedusingYI,

loosepackingDeterminedusing

fluorescencemicrosope

C bead(beads/m

2 )

Appliedbeadconcentration(g/ml)

120 Chapter5

end end %out=tst; mx=mx'; [npks,crap]=size(mx); %if size is specified, then get ride of pks within size of boundary if nargin==3 & npks>0 %throw out all pks within sz of boundary; ind=find(mx(:,1)>sz & mx(:,1)<(nr-sz) & mx(:,2)>sz & mx(:,2)<(nc-sz)); mx=mx(ind,:); end %prevent from finding peaks within size of each other [npks,crap]=size(mx); if npks > 1 %CREATE AN IMAGE WITH ONLY PEAKS nmx=npks; tmp=0.*im; for i=1:nmx tmp(mx(i,1),mx(i,2))=im(mx(i,1),mx(i,2)); end %LOOK IN NEIGHBORHOOD AROUND EACH PEAK, PICK THE BRIGHTEST for i=1:nmx roi=tmp( (mx(i,1)-floor(sz/2)):(mx(i,1)+(floor(sz/2)+1)),(mx(i,2)-floor(sz/2)):(mx(i,2)+(floor(sz/2)+1))) ; [mv,indi]=max(roi); [mv,indj]=max(mv); tmp( (mx(i,1)-floor(sz/2)):(mx(i,1)+(floor(sz/2)+1)),(mx(i,2)-floor(sz/2)):(mx(i,2)+(floor(sz/2)+1)))=0; tmp(mx(i,1)-floor(sz/2)+indi(indj)-1,mx(i,2)-floor(sz/2)+indj-1)=mv; end ind=find(tmp>0); mx=[mod(ind,nr),floor(ind/nr)+1]; end if size(mx)==[0,0] out=[]; else out(:,2)=mx(:,1); out(:,1)=mx(:,2); end

References

1. D.R.Lide, inCRCHandbookofChemistryandPhysics,84thEdition (Taylor&Francis,2003),p.64.

2. M. C. Coen, R. Lehmann, P. Groning, M. Bielmann, C. Galli, and L. Schlapbach,“AdsorptionandbioactivityofproteinAonsiliconsurfacesstudiedbyAFMandXPS,”J.ColloidInterfaceSci.233,180‐189(2001).

3. S. Ohnishi, M. Murata, and M. Hato, “Correlation between Surface Morphology andSurfaceForcesofProteinAAdsorbedonMica,”Biophys.J.74,455‐465(1998).

4. Polysciences, “Technical data sheet Polybead® Polystyrene Microspheres”(Polyscience, Inc. , 16‐05‐2015, 2013), retrieved 07‐09‐2015,http://www.polysciences.com/skin/frontend/default/polysciences/pdf/TDS%20238.pdf.

5. S. N. Kasarova, N. G. Sultanova, C. D. Ivanov, and I. D. Nikolov, “Analysis of thedispersionofopticalplasticmaterials,”Opt.Mater.29,1481‐1490(2007).

6. R.L.Schoch,L.E.Kapinos,andR.Y.H.Lim,“Nucleartransportreceptorbindingaviditytriggersaself‐healingcollapsetransitioninFG‐nucleoporinmolecularbrushes,”Proc.Natl.Acad.Sci.USA109,16911‐16916(2012).

7. I.Bjork,B.A.Petersson,andJ.Sjoquist,“SomephysiochemicalpropertiesofproteinAfromStaphylococcusaureus,”Eur.J.Biochem.29,579‐584(1972).

8. S. Zolls,M.Gregoritza,R. Tantipolphan,M.Wiggenhorn,G.Winter,W. Friess, andA.Hawe,“Howsubvisibleparticlesbecomeinvisible‐relevanceoftherefractiveindexforproteinparticleanalysis,”J.Pharm.Sci.102,1434‐1446(2013).

9. “SPHEROTM Fluorescent Particles”, retrieved 2015‐09‐01,http://www.spherotech.com/fluorescent%20particles%20catalog%202010‐2011%20rev%20a.pdf.

Chapter6

Size‐selectiveanalytedetectionusingmultiplewavelengthsandpolarizations

AbstractUntil now, size‐selective detection with the YI sensor was done using multiplewavelengthsbasedontheirdifferentmodeprofilesandtheirdifferentsensitivitiesindifferentlayersintheevanescentfield.Chapter6showsthatinasimilarway,multiplepolarizationscanalsobeusedtodiscriminatebetweenrefractiveindexchanges( n’s)ofmultiplelayersormultipledifferent‐sizedsubstances.Calculationsfromchapter2 showed that themore different the sensitivity coefficients of eachmode, the lesssingularthematrixtosolve n ’sinmultiplelayersorinducedbymultiplesubstances,meaning that all noise sources in effN show up less enhanced.We show that it is

possibletomeasuremultiplepolarizationsandmultiplewavelengthssimultaneously,resultinginamatrixwhichislesssingularforsomecasesandthereforerepresentingan improvement of the approach in these cases.However, it shouldbenoticed thatmeasuring multiple wavelengths and polarizations also lead to an increase of theartefacts(duetoboundaryeffectsoftheshiftinginterferencepatternorduetogratingor lens aberrations in the imaging path of the setup) in the measurements. For asuccessful applicationof theuseofmultiplepolarizationsandwavelengths for size‐selectivedetection,theartefactsinthemeasurementsshouldbereducedsignificantly.

122 Chapter6

6.1Introduction

WehavedemonstratedinChapter5thatsize‐selectivediscriminationofanalytescanbedoneusingmultiplewavelengths.However,suchdiscriminationcanalsobedoneusing different polarizations or a combination of wavelengths and polarizations,because different polarizations also have different mode profiles and thereforedifferentsensitivities indifferent layers in theevanescent field. It isalsopossible todetermine both the thickness and density (RI) of a protein adlayer using multiplepolarizations, as was illustrated by Cross et. al [1] and Swann et. al [2] using dualpolarization interferometry. Calculations from chapter 2 showed that wavelengthsshould be chosen widely spread such that the sensitivity coefficients of eachwavelengthdifferthemost.Consequently, thematrixtosolve n inmultiplelayersor inducedbymultiplesubstances is lesssingular,meaningthatallnoisesources in

effN willbelessenhanced.Therefore,usingmultiplepolarizations,eitheratoneorat

multiple wavelengths, results in a less singular matrix, thus leading to animprovementintheexperimentalrealization.

To determine a N number of n ’s in multiple layers or induced by multiple

substancesfromaNnumberof effN ’s(measuredatmultipledifferentwavelengths

and/orpolarizations),usinganyanalysisapproach,a N N matrixisrequired.Toseehow well the solution of n can be determined from the matrix equation, the

conditionnumber ofthematrixcanbedetermined.Foralinearsystem Ax b ,the

condition number gauges the transfer of error from matrix A and vector b into

vector x [3].Forasmallconditionnumbertheproblemiswell‐conditioned,whereasa large condition number corresponds to an ill‐conditioned problem. The rule of

thumbis thattheexpected loss insolvingaproblemof Ax b isat leastkdigitsof

precision for a condition number ( ) 10kA [3]. When the determined condition

numberofthematrixislowerusingthemultiplepolarizationsandwavelengths,this

will improvethesetupastheerrorsin n and S willbelesstransferredinto effN .

However, the implementation of the use ofmultiple polarizations andwavelengthsrequiresextracomponents in thesetup tomeasure the transverseelectric (TE)andtransversemagnetic(TM)polarizedlightindependently.Therefore,theimprovementof precision of n should be weighed against the drawbacks (e.g. extra costs,complexityoftheinstrumentalsetup,alignmentoverhead).Toillustratethefeasibilityofmeasuringsimultaneouslymultiplewavelengthsandpolarization,theexperimentalrealisation and proof‐of‐principle experiments are shown in section 6.2 and 6.3respectively.The chapter endswith a general discussion and conclusions in section6.4.

Size‐selectiveanalytedetectionusingmultiplewavelengthsandpolarizations 123

6.2Experimentalrealisation

The setupwith butt‐end coupling using a SM‐PM fiber as described in Chapter 3 isusedfortheexperimentsshowninthischapter.Thefiberwasrotated45degreestocouple both TE and TMmodes into the waveguide. After coupling out of the YI, aWollaston prismwas used to split up the polarizations to be able to independentlydetecttheinterferencepatterns.InordertodeterminetherequiredsplittingangleoftheWollastonprismwhichispositionedafterthefirstcylindricallensandbeforethegrating,theformulaofatransmissiongrating[4]isrequired:

sin sinm ia m , (6.1)

whereaisthedistancefromthecentreofoneslittothecentreoftheadjacentslit, m

the angle of the mth maximum and i the angle of the incident beam which is

determined by theWollaston prism. This can be combinedwith the formulawhichdescribestheheightofthefocus h asafunctionoftheangleoftheincidentbeamatthelens m ,whichisgivenbytheexpression:

tan mh f , (6.2)

wherefisthefocaldistanceofthelens.Themaximaldistancebetweentheoutermostinterferencepatternsshouldbesmallerthan256x26μm=6.656mm,which is thesizeoftheCCDchip.Combiningformula6.1and6.2andassumingf=50mm,andthewavelengths 457, 561 and 660 nm, a sufficient incident angle i which should be

realised by theWollaston prismwas determined to be 1º. Therefore, a PWQ 60.25Wollastonprism(BernhardHalleNachfl.GmbH,Berlin)madeofquartzcrystalwithadivergenceangleof1º@550nmwithaclearapertureof24.5mmwasplacedintothesetupinbetweenthefirstcylindricallensandthetransmissiongrating.

Fig.6.1:CCDimageoffourinterferencepatternscorrespondingtoTEandTMpolarizedlightof =457nmand561nm.

124 Chapter6

Fig. 6.1 shows four interference patterns of TE457, TM457, TE561 and TM561imagedontheCCD(notethatthecoloursareartificialandcorrespondtothelookuptableused).Itwasalsopossibletomeasurephasechangesofsixdifferentinterferencepatternsdodiscriminatebetweenthreesubstances,butthisdoesnotimprovetheendresult very much compared to measuring the phase changes of three differentinterference patterns (see Chapter 2). Therefore, it was chosen to image only twowavelengths with two polarizations such that these four interference patterns fallover large number of pixels rows but still do not overlap such that they can bemeasuredindependentlyandwiththelowestpossiblephasenoise.

6.3Proof‐of‐principleexperiments

ProteinAandD‐glucoseTo illustrate if two substances can be discriminated from each other using acombination of different wavelengths and polarizations, an experiment was donewhere6.16mg/mlD‐glucoseand1μg/mlproteinAwhereaddedsimultaneouslyto

thesensorasshowninFig.6.2a.The effN wasmeasuredovertimeforTEpolarized

lightat =457nmand660nmsimultaneouslywithTMpolarizedlightat =457nm. First, a 1x PBS buffer is completely replaced by D‐glucose which is after 500s

completelyreplacedbyPBSagain,illustratedbythe effN levelwhichcomesbackto

the original level before adding the D‐glucose. Next, D‐glucose is addedsimultaneouslywith protein A and again 500s later awashing step is applied. The

effN doesnot goback to its original level, becauseproteinA is left at the surface.

Afterthewashingstep,thereisaslightdecreaseofthe effN signalwhichisprobably

becausedbydesorptionoftheproteinA.

Threedifferentcombinationsoftwoofthethreemeasured effN ’swereusedto

determine n for the D‐glucose and the protein A and shown in Fig. 6.4b‐d. Theanalysiswasdonewiththeratio‐basedapproach.Themeasuredratioswerefittedtothe theoretical model to determine a thickness and consequently determine theamplitudeofthesignal(combinedapproach).FortheTMmodes,themeasuredratiosfor the protein A were very different from the theoretical ratios. Therefore, thetheoretical model used to determine sensitivity coefficients in Chapter 2 [5] wascomparedwithaneffectiveindexapproximationbasedmodesolverprogram[6].The

correctnessoftheusedtheoreticalmodeltodetermine effN wasverifiedasthe effN

’s of TE and TMmodes gave the same values as an effective index approximationbasedmodesolverprogram.Thesensitivitycoefficientswerealsodeterminedbythe

samemodesolverprogram,bycalculatinga effN ,subsequentlyinducinga n (ina


layeror inbulk), thencalculatingagain the effN , anddividingthedifference in the

two effN ’sbytheinduced n .ThevaluesforthesensitivitycoefficientforTEwhere

approximately (<0.1% for bulk and <1% for layers) the same compared to thesensitivity coefficients based on the theory in Chapter 2. However, the sensitivitycoefficients of TM did not correspond well. For the bulk the deviation was small(<1%),butfordifferentlychosenlayersthedeviationvariedfrom19‐27%.Therefore,we did not use the same model as in Chapter 2, but we used the effective indexmethodtodeterminetheoreticalratiosandcomparethemwiththedeterminedratios.

From the measurement we determine the ratios of the differentwavelengths/polarizationsby fitting themeasured data to the expected values (seeChapter 4). The proteinn should be approximately zerowhen adding D‐glucose and

that glucosen should be approximately zero after the second washing step

(approximatelyzeromeansinthiscasethatthevalueshouldbeinlinewiththevalueexpectedbasedon lineardrift).Basedon thisassumptions, theratioswerechangedand this resulted for TE457with TE660 in optimal ratios of glucose

660/ 457TE TER = 1.20 andprotein660/ 457TE TER =0.69,whichcorrespond toadcore=62.0nmandadprotein =2.0nm.For

TE457andTM457thefittingwiththeratio‐basedapproachresultedinoptimalratiosof glucose

457/ 457TM TER =1.44and protein457/ 457TM TER =1.06,correspondingtoadcore=66.5nmanda

dproteinwhichcouldnotbefitted(ratiofor2.1nm=1.075).FittingbasedonTM457andTE660 resulted ratios of glucose

660/ 457TE TMR = 0.825 and protein660/ 457TE TMR = 0.65, agreeingwith a

dcore=70.5nmandadprotein=24.0nm(ratiofor2.1nm=0.618).Thevaluesfordproteinare not realistic. The differences might be explained by the artefacts in themeasurement (due to boundary conditions of the shifting interference pattern, seeChapter3)whicharemuchlargerforTE457comparedtomeasurementswithouttheWollastonprisminthesetup(seeFig.6.2e/f).Theartefactsaredifferentinamplitudefor each polarization/wavelength and therefore probably also different in the

measuredvaluesof effN foreachpolarization/wavelength,whichresultsindifferent

ratiosbetweenthem.Thiscanalsobeseeninthestrongartefactsin n (seeFig.6.2d)which were not observed in the analysed data of the measurements without theWollaston prism in the setup. This illustrates that the artefacts are present to agreaterextentwiththeWollastonprisminthesetup.Duetotheincreasingartefactsinthemeasurements,itwasnotpossibletodiscriminatetheproteinAcompletelyfromtheD‐glucosewhenaddingthemsimultaneouslywhenanalysingtheresultsonTE457andTE660orTM457andTE457.TheD‐glucoselevelistoohigh(shouldbethesameas the firstD‐glucose step)and theproteinA showsanunexpecteddrop.When theartefactsare

126 Chapter6

Fig. 6.2: a) effN over time when adding 6.16 mg/ml D‐glucose alone and together with 1

μg/ml,b,c,d)theanalysedsignalof n forTE457/TM457,TE457/TE660andTM457/TE660respectively,alldeterminedwiththeratio‐basedapproach,e)themeasuredrelativeamplitudeandf)thetypicalrelativeamplitudeforameasurementwiththeTE457andTE660withouttheWollastonprisminthesetup.

0 500 1000 1500 2000 2500

‐1

0

1

2

3

4

0 500 1000 1500 2000 2500

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

0 500 1000 1500 2000 2500

0

4

8

12

0 500 1000 1500 2000 2500

0

4

8

12

16

0 500 1000 1500 2000 2500

0

4

8

12

0 500 1000 1500 2000 2500

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Neff

Time(s)

TE660TM457TE457

x10‐4


PBS

PBS

PBS

6.16mg/mlD‐glucose+1g/mlproteinA

Normalizedamplitude

Time(s)

TE660TM457TE457

f

x10‐5

n u

nscaled

Time(s)

ProteinAD‐glucose

TE457/TM457d

b

e

cTE457/TE660x10‐5

n u

nscaled

Time(s)

ProteinAD‐glucose

aTM457/TE660x10‐5

n u

nscaled

Time(s)

ProteinAD‐glucose

Normalizedamplitude

Time(s)

TE660TE457

TypicalmeasurementwithoutWollastonprism


lower, in the case of TM457 andTE660, itwas possible to sufficiently discriminatebetween protein A and D‐glucose, illustrating that it is possible to do thediscriminationbasedonTEandTMmodesofdifferentwavelengths.

Theconditionnumberofthematrixusedwiththeratio‐basedapproachwerealsodetermined.The ’swere13.5,7.65and17.8forTE457/TM457,TE457/TE660andTE660/TM457 respectively. This showed that in this case the combination ofpolarizationsandwavelengthswasnotanimprovementcomparedtomeasuringwithTE457andTE660nmonly,whichshows thatwavelengthsandpolarizationsshouldbechosencarefully.85nmbeadsandD‐glucose

To discriminate binding of 85 nm beads from a D‐glucose bulk change, effN was

measuredatTE457,TE561andTM561whenadding6.16mg/mlD‐glucoseonlyandtogetherwith1 μg/ml85nmbeads as shown inFig. 6.3a. Fig 6.3 b shows that theamplitudesoftheartefactsinthemeasurementsareagainrelativelyhigh(upto40%forTM561).Theartefactsshowuppredominantlyintheanalysedsignalof n basedon TM561 (see Fig. 6.3b‐d),which is expected as the ’s of thematrices used are18.95, 61.35 and 13.74 for TE457/TM561, TE561/TM561 and TE457/TE561respectivelyand theartefactsare thehighest forTM561.ComparingFig.6.3bandcthe artefacts in the end result in n based on TE561/TM561 are clearly higherdespite comparable normalized amplitudes of TE457 and TE561 (see Fig. 6.3b),illustratingtheeffectofthehigherconditionnumberofthematrixofTE561/TM561.

Theratios in effN wereagain fittedasstatedbeforeandare glucose561/ 457TM TER =1.025,

561/ 457beadTM TER =0.828, glucose

561/ 561TE TMR =1.091, 561/ 561beadTE TMR =1.022, glucose

561/ 457TE TER =1.125 and

561/ 457beadTE TER =0.950. The ratios were fitted to the effective indexmethod resulting in

dcore=66.6nm,68.0nmand58.2nm,andadbead=97nm,anunfittablevalueand62nmcorrespondingtoTM561/TE457,TE561/TM561andTE561/TE457respectively.However, because of the large artefacts in themeasurements these values are veryunreliable and not usable. Artefacts should be reduced significantly for accuratediscriminationbetweensubstancesanddeterminationofsubstancethicknesses.

128 Chapter6

Fig. 6.3: a) effN over time when adding 6.16 mg/ml D‐glucose alone and together with 1

μg/ml 85 nm beads, b) the corresponding signal of the relative amplitude and c, d, e) theanalysedsignalof n forTE561/TM561,TE561/TM457andTM457/TM561respectively,alldeterminedwiththeratio‐basedapproach.

0 500 1000 1500 2000 2500‐2.5

‐2.0

‐1.5

‐1.0

‐0.5

0.0

0 500 1000 1500 2000 25000.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

0 500 1000 1500 2000 2500

‐4

‐2

0

2

0 500 1000 1500 2000 2500‐2.0

‐1.5

‐1.0

‐0.5

0.0

0.5

1.0

0 500 1000 1500 2000 2500‐4

‐2

0

x10‐4Neff

Time(s)

TE561TM561TE457

Normalizedamplitude

Time(s)

TE561TM561TE457

PBS

PBSPBS6.16mg/ml

D‐glucose

6.16mg/mlD‐glucose+1g/ml85nmbeads

TE457/TM561x10‐5

n u

nscaled

Time(s)


TE561/TM561x10‐5n u

nscaled

Time(s)


e

dc

ba

TE457/TE561x10‐5

n u

nscaled

Time(s)



ProteinA,85nmbeadsandD‐glucoseTo illustrate the possible discrimination of three substances based on multiplepolarizationsandwavelengths,proteinA,85nmbeadsandD‐glucosewerealsoaddedto sensor within one measurement as shown Fig. 6.4. First, at t1 only D‐glucose isaddedtothesensorandatt2awashingstepisapplied.Next,att385nmbeadswereaddedtothesensorafteragainapplyingawashingstepatt4.Betweent5andt6proteinAwasaddedtothesensor.Next, n wasdeterminedwiththeratio‐basedapproach.The ratioswere again fitted to the expected values of n , resulting in 457/ 457

proteinTM TER

=1.045, 561/ 457proteinTM TER =0.53, 457/ 457

beadTM TER =1.39, 561/ 457

beadTM TER =0.819, glucose

457/ 457TM TER =1.465,glucose

561/ 457TM TER =1.017.Theseratioscorrespondtodcore=67.0nm,66.5nmand65.8nm,an

unfittabledprotein(2.1nmcorrespondsto 457/ 457proteinTM TER =1.091, 561/ 457

proteinTM TER =0.563and

561/ 457proteinTM TMR = 0.516) and dbead = 131 nm, 110 nm and 69 nm for TE457/TM457,

TE457/TM561andTM457/TM561respectively.Thedifferencesbetweenmeasured

Fig.6.4:a) effN overtimewhenadding6.16mg/mlD‐glucose(t1‐t2),1μg/ml85nmbeads(t3‐

t4)and2μg/mlproteinA(t5‐t6)tothesensor,allmeasuredatTE457,TM457andTM561,b)thecorresponding signal of the relative amplitude, c) the analysed signal of n using the ratio‐basedapproach,d) effN overtimewhenadding6.16mg/mlD‐glucose(t1‐t2),1μg/ml85nm

beads (t3‐t4) and 2 μg/ml protein A (t5‐t6) to the sensor, allmeasured at TE457, TE561 andTE660,e)thecorrespondingsignaloftherelativeamplitudeandf)theanalysedsignalof n usingtheratio‐basedapproach.

0 1000 2000 3000

0

1

2

3

4

5

0 1000 2000 30000.6

0.8

1.0

1.2

1.4

0 1000 2000 3000

0

2

4

6

8

0 1000 2000 3000

0

1

2

3

4

5

0 1000 2000 30000.6

0.8

1.0

1.2

1.4

0 1000 2000 3000‐1

0

1

2

3fed

t6

t5t4

t3

x10‐4

Neff

Time(s)

TM561TM457TE457

t1

t2

Normalizedamplitude

Time(s)

561nm,TM457nm,TM457nm,TE

x10‐5

nunscaled

Time(s)

ProteinA85nmbeadsD‐glucose t6

t5

t4t3

t1

t2

t6

t5t4

t3

x10‐4

Neff

Time(s)

660nm,TE561nm,TE457nm,TE

t1

t2t6

cba

Normalizedamplitude

Time(s)

457,TE561,TE660,TE

x10‐5

nunscaled

Time(s)


t5

t4t3

t1

t2

130 Chapter6

and theoretical values are probably again explained by the artefacts in themeasurements.Despitethesemeasuredvaluesdonotmatchtheexpectedthicknessof≈2.1nmfortheproteinand85nmforthebeads,itwasstillpossibletodiscriminatebetween the three independent sources using the ratio‐based approach, illustratingthestrengthoftheratio‐basedapproach.Theoreticalparametersarenotrequiredforthe ratio‐based approach and just normal calibrationmeasurements canbeused todeterminetheratiosoftheindividualsubstances.

Thematrix basedon the determined ratio has a =68.94,which ismore thanthree times lower than thematrix basedonmeasurementswithTE457,TE561 andTE660,which had a = 209.41. The artefacts in the relative amplitude in the casewhen measuring with multiple wavelengths and polarizations is higher (standard

deviation is 28.74 10 ) compared to measuring with only multiple wavelengths

(standarddeviation is 23.75 10 )asshownFig.6.4b/e.However, theartefacts thatshowupin n arecomparable(seeFig.6.4c/f),astheartefactsarelessenhancedforthemultiplewavelengthsandpolarizationsbecauseofthelowerconditionnumberof

thematrixwhichisusedtodetermine sn from effN .AttheadditionofD‐glucose,

theartefactsarerelativelylowfor proteinn and beadsn forthepolarization/wavelength

case. On the other hand, the artefacts in glucosen are a bit higher for the

polarization/wavelength case. So, when lowering the artefacts in effN when

measuringwithmultiplewavelengthsandpolarizations it ispossibleto improvethesetup as the artefacts are less enhanced in this specific case due to the betterconditionedmatrix( is3xlower).


UsingdifferentwavelengthsandpolarizationsitwaspossibletodiscriminatebetweenD‐glucose,85nmbeadsandproteinA.However, thepolarizationsandwavelengthsshould be chosen carefully when they are used for improvement of the setup.Choosing the wrong polarization/wavelength combination results in a higher resultinginmoreenhancementofallnoisesources.MeasuringwithTE4567,TM457andTM561wewereabletoreducethe threefoldcomparedtoTE457,TE561andTE660, corresponding to an approximately 4.8 times lower enhancement of noise.Moremeasurementsortheoreticalcalculationsarerequiredtodetermineifthereareevenmore favourable combinationsofwavelengths andpolarizations to reduce the ,whichstronglydependsonthemeasurablesubstances,evenmore.

Despite the decrease of , the artefacts due to boundary effects of the shiftinginterference patterns were increased using a Wollaston prism to split up theinterference patterns, illustrating the optical components influence artefacts.


Therefore,thedeterminedsignalsof sn showednoisysignalsdespitethedecrease

of . Consequently, the artefacts should be strongly reduced for a successfulapplication of the size‐selective detection using multiple wavelengths andpolarizations, using for example larger optical components and detectingmore andlargerfringesofthe interferencepattern.Alternatively, it ispossibletoalternatinglymeasureTEandTMmodesbyswitchingbetween thesemodesusinga ferroelectricliquidcrystal1/2waveplate[1].

Themeasuredratiosbetween effN ofdifferentwavelengthsandpolarizationsdo

not match with the theoretical expected ratios, especially for the TM modes.Furthermore, it seems that the theory used in Chapter 2 [5] for determination ofsensitivitycoefficientsforTEmodesfordifferentregionsintheevanescentfieldisnotapplicable forTMmodes as there is a large (up to almost 30%)mismatchwith theeffectiveindexmethodfromHammer[6].Therefore,theapproximationtodeterminethesensitivitycoefficientsper layerbySchipper[5]shouldnotbeappliedusingTMmodes.FittingthemeasuredratiostothetheoreticaleffectiveindexmethoddidleadtofittedratiosfortheproteinAandbeadswhichwerenotreliablebecauseofthevery

high artefacts in themeasured effN . Next to the influence of artefacts on the end

result of sn , it might be possible that some light is converted on the chip. Light

convertedfromTEtoTMmodesortheotherwayaroundcouldnotbefilteredoutasthe polarizer had to be removed from the setup to be able to measure twopolarizations. To determine if light was converted from TE to TMmodes, separatemeasurements with only TM light with polarizer in imaging path are required tocomparethemwiththeresultsofthecombinedTE/TMmeasurements.MeasurementswithTE light showcomparable resultswith thecombinedTE/TMmeasurements. Itshould be noted that when only measuring with TE or TM light, modes that areconvertedtwice(e.g.TEtoTMandbacktoTE)cannotbefilteredout.Forsuccessfulapplication the artefacts should be strongly reduced. Subsequently, repeatabilitystudiesare required todetermine ifmeasuredratiosareconstantwhenartefacts in

effN arereduced.

Notwithstanding themismatchbetween theory andmeasured ratios, itwas stillpossibletodiscriminatebetweenthethreeindependentsourcesusingtheratio‐basedapproach. Theoretical parameters are not required for the ratio‐based approachillustrating the strength of the ratio‐based approach. However, for a successfulapplication of the use of multiple polarizations for size‐selective detection, theartefactsinthemeasurementsshouldbereducedsignificantly.

132 Chapter6

Acknowledgements


References

1. G. H. Cross, A. A. Reeves, S. Brand, J. F. Popplewell, L. L. Peel,M. J. Swann, andN. J.Freeman,“Anewquantitativeopticalbiosensorforproteincharacterisation,”Biosens.Bioelectron.19,383‐390(2003).

2. M. J. Swann, L. L. Peel, S. Carrington, and N. J. Freeman, “Dual‐polarizationinterferometry: an analytical technique to measure changes in protein structure inreal time, to determine the stoichiometry of binding events, and to differentiatebetweenspecificandnonspecificinteractions,”Anal.Biochem.329,190‐198(2004).

3. E.CheneyandD.Kincaid,NumericalMathematicsandComputing (CengageLearning,2007).

4. E.Hecht,“Optics,”4ed.(AddisonWesley,2002),p.477.5. E. F. Schipper, “Waveguide immunosensing of small molecules,” Ph.D. Thesis,

UniversityofTwente(1997).6. M.Hammer, “1‐Dmultilayerwaveguidemodesolvereffective indexapproximation”

(2015),retrieved23‐07‐2015,http://www.computational‐photonics.eu/oms.html.

Chapter7

TimingdifferenceofrefractiveindexchangesinducedbybindingofproteinsandbulkchangesAbstractFromthedetermined n ’susingsize‐selectivedetectionwesawthatthe n duetobindingofproteinAtothesurfaceoccurredearlierthanthe n duetoproteinAinthebulk(afewhundredsofnanometresabovethesensorssurface).However,itwasexpectedthatproteinAshouldarriveinthebulkbeforeincanattachtothesurface.InChapter7thisunexpectedtimedelayisdiscussed.Finiteelementmethodsimulationson the diffusion and binding ofmolecules show that a time delay between surfacebindingofproteinAandproteinAbulkchangescanarisewhenproteinAhasahighaffinity for the surface.Thebulk region isdepleteddue to the fact thatproteinA isstrongly attracted towards the surface. The timedelay increaseswith an increasingmaximum surface capacity max and increasing adsorption rate ka. For a ka of 105

m3/mol/s and a max between 4.4x10‐3 and 1.7x104 fg/mm2 (corresponding to a

protein A radius of 1 and 2 nm respectively which agrees well to the measuredthicknessof2.1nm(seeChapter4)),a timedelayofapproximately20s to100s isfound. This is comparable to the observed time delay. To verify the results of thesimulations,proteinAandD‐glucosewereaddedsimultaneouslytothesensor.DuetothefactthatD‐glucosehasamuchhigherdiffusionthanproteinAitwasexpectedthatn due to D‐glucose bulk changes would appear earlier now. However, again the

signal of n due to binding of protein A occurred earlier than the n due to D‐glucose.Experimentshintthatthistimedelayisrelatedtothetransporttowardsthesensingwindowas increasing the transporting tube lengthresulted inan increasingtimedelay.Experimentsshowedthatthe timedelayalsooccurred forbovineserumalbumin,suggestingthatthetimedelayisindependentoftheproteinused.However,theexactoriginofthetimedelaybetweenD‐glucosebulkchangesandsurfacebindingofproteinAisnotfound.

134 Chapter7

7.1Observedtimedelayofbulksignalcomparedtobinding

7.1.1ProteinAbindingearlierthanproteinAbulkProtein A andD‐glucosewere added separately from each other to the sensor. Fig.7.1a,b show the effN over timemeasured at 457, 561 and 660 nm of two similar

experiments and Fig. 7.1c‐f the corresponding n determined for D‐glucose andproteinAusingthecombinedapproach(seeChapter4).TheproteinAwasaddedinconcentrationsintheorderof100μg/ml,resultingintwomeasurableeffects:asignalinthefirst≈2nmduetothebindingofproteinAtothesurfaceandabulksignalintheregionof≈200nmduetotheproteinAinsolution.Bothsignalshaveadifferentslopewhich is already seen in themeasured signal of effN in Fig. 7.1a, b, but this

showsupmoreclearlyinthedeterminedsignalof n (seeFig.7.1e,f),inwhichitwasobservedthatthesignalduetobindingofproteinAstartedearlierthanthebulksignalduetoproteinA.ThebulksignalduetothehighconcentrationofproteinAresultsinthe same ratios of effN as the D‐glucose bulk signal, whichmeans that both bulk

signalscouldbeanalysedusingasingleratioforbulkeffects.

7.1.2ProteinAbindingearlierthanD‐glucosebulkAlso experiments were done where protein A and D‐glucose were addedsimultaneously to thewaveguide.Fig.7.2a,bshowthe effN over timemeasuredat

457,561and660nmoftwoexperimentswhereD‐glucoseandproteinAwereaddedsimultaneously to the waveguide and Fig. 7.2c‐f shows the corresponding n determined for thebulk signaland theproteinAusing the combinedapproach.Forthese experiments a lower protein A concentration was used, such that the bulkcontribution due to the protein A can be neglected (checked when protein A wasaddedalonetothesensorintheselowconcentrations)andthatthebulkcontributionisonlycausedbytheD‐glucose.AlsofortheseexperimentsthesignaltothebindingofproteinAtothesurfacestartsearliercomparedtothesignalduetothebulkeffectofD‐glucose.Fig.7.2e,fshowthatthisdelayislongerforproteinAconcentrationof26.7μg/ml compared to 1 μg/ml. However, because of the artefacts due to boundaryeffectsoftheshiftinginterferencepattern(seeChapter3),itisdifficulttodetermineexactlywhenthesignalstartsincreasingandthereforeitisdifficulttodeterminetheexacttimedelay.Furthermore, thetimedelaycouldbeanartefact fromthemethod.To answer these questions, a timing experiment is done where protein A and D‐glucosewereaddedtothewaveguideatthesametime,butindifferentchannels.Fig.7.3showstheresultsofthismeasurement.First,atimingstepofwaterto

TimingdifferenceofRIchangesinducedbybindingofproteinsandbulkchanges 135

Fig.7.1:a,b)Twoexperimentswhereshowing effN overtimeat457,561and660nmadding

proteinAandD‐glucose separately to thewaveguide, c,d) the corresponding n due to thebulkandthebindingoftheproteinsdeterminedwiththecombinedapproachande,f)azoom‐inofc,d).

0 1000 2000 3000

0

2

4

6

0 1000 2000 3000

0

2

4

6

0 1000 2000 3000

0

2

4

6

8

10

12

0 1000 2000 3000

0

2

4

6

8

10

12

900 1000 1100 1200 1300

0

2

4

6

8

10

500 600 700 800 900

0

2

4

6

8

10

PBS

133g/mlproteinA

Neff

Time(s)

660nm561nm457nm

PBS


x10‐4

PBS PBS


80g/mlproteinA

fe

dc

ba x10‐4

Neff

Time(s)

660nm561nm457nm

PBS

PBS

n p

roteinA

Time(s)

BindingproteinA(1.9nm)

0

2

4

6

8

10

12dcore=68.5nm

Bulk

n b

ulk

x10‐2 x10‐4n p

roteinA

Time(s)


0

2

4

6

8

10

12dcore=68.5nm

Bulk

n b

ulk

x10‐2 x10‐4

n p

roteinA

Time(s)


0

2

4

6

8

10dcore=68.5nm

Bulk

n b

ulk

x10‐2 x10‐4

n p

roteinA

Time(s)


0

2

4

6

8

10dcore=68.5nm

Bulk

n b

ulk

x10‐2 x10‐4

136 Chapter7

Fig.7.2:a,b)Twoexperimentswhereshowing effN overtimeat457,561and660nmadding

proteinAandD‐glucosesimultaneouslyto thewaveguide,c,d) thecorresponding n duetothebulk and thebindingof theproteinsdeterminedwith the combinedapproachande, f) azoom‐inofc,d).

0 500 1000 1500

0

2

4

6

0 500 1000 1500 2000 2500

0

1

2

3

4

5

0 500 1000 1500

0

2

4

6

8

10

12

0 500 1000 1500 2000 2500

0

1

2

3

4

300 400 500 600 700

2

4

6

8

10

12

14

1200 1300 1400 1500 16004

6

8

10

12

14

16

x10‐4Neff

Time(s)

660nm561nm457nm

26.7g/mlproteinA+6.16mg/mlD‐glucose

PBS

PBS

x10‐4

Neff

Time(s)

660nm561nm457nm


PBS


PBS

PBS

n p

roteinA

Time(s)

BindingproteinA(2nm)

0

2

4

6

8

10

12bulk

n D

‐glucose

dcore=69nmx10‐2 x10‐4

n p

roteinA

Time(s)


0

2

4

6

8

10

12bulk

n D

‐glucose

x10‐2 x10‐4dcore=65nm

n p

roteinA

Time(s)


‐1

0

1

2

3

4

bulk

n D

‐glucose


n p

roteinA

Time(s)


‐1

0

1

2

3

4fe

dc

ba

bulk

n D

‐glucose



phosphate buffered saline (PBS) is done, to cancel out timing differences due todifferenttubelengthsbelongingtothedifferentchannels.Thetimingdifferenceswerecompensated for by shifting the signal by thedelay found for thewater ‐ PBS step.After compensation of time differences due to tube length, also a time delay of D‐glucose compared to the binding of protein A is observed. Therefore, it can beconcludedthatthemeasuredtimedelayofbulkeffectscomparedtobindingisnotanartefactoftheanalysismethodbutisreal.

But why do we see a time delay of bulk signal due to protein A or D‐glucosecomparedtobindingofproteins?Intuitivelyonewouldsaythatthemoleculesshouldarrive first in the bulk (≈ 200 nm above the surface) before they can bind to thesurface.Moreover,assuminglaminarflow,thetransportofmoleculesnearthesurfaceis determined by diffusion of molecules as the velocity is nearly zero. Glucosemoleculesdiffusefaster(Dg=6.7x10‐10m2/s[1])thanproteinAmolecules(Dp=4.3x10‐11m2/s[2])andtherefore,theglucosemoleculesshouldarriveearlieratthebulkregion than theproteins reach the surface. SimulationswithCOMSOLweredone tofindanexplanationforespeciallythetimedelayofproteinbulksignalwithrespecttothebindingofproteinA,butalsotheD‐glucoseisconsideredinthemodel.ResultsofthisCOMSOLsimulationarepresentedinsection7.2.

Fig. 7.3: a) Timing experiment showing effN over time at 660 nm measured at different

channelstowhichproteinAandD‐glucosewereaddedseparatelyandb)azoom‐inofa).

7.2SimulationsD‐glucoseandproteinAinsensingwindow

The theory in this section is based on research on convection, diffusion andadsorptioninsurface‐basedbiosensorsofHansenet.al[3].Weslightlychangedtheir

0 1000 2000 3000

0

2

4

6

8

2100 2200 2300 2400 25003

4

5

6

pbs

pbs

H2O

H2Oin

allchannels

x10‐4

Neff

Time(s)

660nm,channel2660nm,channel3660nm,channel4

PBSinchannel2,3,4

PBS


26.7g/mlproteinA

H2O

b x10‐4

Neff

Time(s)

660nm,channel3660nm,channel4

a

138 Chapter7

model, because their research only considered unimolecular systems. We study amodelwithbothproteinsandglucoseinsolution.COMSOLMultiphysics4.3wasusedforsimulationsandamodelcalled‘transportandadsorption’wasusedasabasisforourmodel.

7.2.1SystemgeometryAflowchamberisplacedontopofthesensingwindowsofthechip.Thedimensionsoftheflowchamber,includingthesensingwindow,areshowninFig.7.4.Thescalesofthe flow chamber are l = 6mm,w = 2mm,h = 100 μm. The length of the sensingwindowlsis4mmandthewidthofthesensingwindowwsis4μm.Theflowthroughtheflowchamberisinx‐directionandbecauseofthelowReynoldsnumbertheflowislaminar.Forthatreasonandbecauseofthelargeaspectratiow/h≈15(sidewallsinz‐direction are expected to not influence the sensing window), invariance in z‐directionisassumedandconsequentlythesystemcanbedescribedbya2Dmodel[3]as shown in Fig. 7.4. The velocity of the laminar flow vlam can be described by thefollowingequation:

max4 1lamv v y h y h (7.1)

wherevmaxisthemaximumvelocity.

Fig.7.4:Flowchamberandsensingwindowdimensionsina3Dand2Dconfiguration,wherehis the height of the flow chamber,w thewidth of the flow chamber, l the length of the flowchamber,ls,thesensingwindowlength,wsthesensingwindowwidthandvlamthevelocityofthelaminarflow.

7.2.2TheoreticalequationsThree important variables that are studied are defined as: cg, which is the averageconcentrationofD‐glucoseinaregionof200nmhighabovethesensingwindow;cp,whichistheaverageconcentrationofproteinAinaregionof200nmhighabovethe


sensing window; and , which is the average protein A surface coverage on thesensingwindow.First,thespatiotemporaldevelopmentofcgcanbedescribedby:

2g glam g g

c cv D c

t x

, (7.2)

where2 22 2 2x y is the Laplacian and Dg is the diffusion coefficient of

glucose.Forglucosetheassumptionismadethatthereisnoadsorptionofmoleculesat thewalls of the flowchamber,whichmeans that there is a non‐flux condition atthosenon‐adsorbingdomains,whichisgivenby:

0gc

y

. (7.3)

Second,thespatiotemporaldevelopmentofcpisdescribedbyasimilarequation:

2p plam p p

c cv D c

t x

, (7.4)

whereDpisthediffusionequationoftheproteins.Theboundaryequationsaregivenby:

0pc

y

, (7.5)

whichisano‐fluxconditionwhichholdsforthenon‐adsorbingregions(flowchamberwallsexceptforthesensingwindow)and

,pp p

cD R c

y

(7.6)

which is a balance between diffusive flux perpendicular to the surface and theadsorptionrate(whichisgivenbyequation(7.9))andholdsforthesensingwindow,towhichproteinscanadsorb.

Third,thespatiotemporaldevelopmentof isdescribedbyadsorption‐diffusionequation:

,s pD R ct x x

, (7.7)

whereDsisthesurfacediffusioncoefficientoftheproteinsand , pR c describesthe

netrateofchangeof ,duetoadsorptionanddesorptionofmolecules.Moleculescanonlyleavethesensingwindowbydesorption,whichresultsinthefollowingboundaryconditionatbothendsofthesensingwindow:

140 Chapter7

0x

. (7.8)

Finally, theadsorptionratecanbedescribedbytheLangmuirmodel. It isa firstorder scheme between bulkmolecules at the interface

0yc

and free surface sites

max ,andcanbedescribedby:

max0, p a dy

R c k c k

, (7.9)

wheremaxisthemaximumsurfacecapacity,katheadsorptionrateconstantandkdthedesorptionrateconstant.Thekaandkdvaluesareassumedtobeindependentofthesurface density. At high density it could be expected that there will be interactionbetween the bound molecules. Nevertheless, this model can explain in general themajorityofadsorptionanddesorptionprocessesinmolecularbiology[4].

7.2.3ParametervaluesIn this sectionwe summarize all the parameters and their values andwemake anestimationoftheunknownvalues,whichwillallbeusedinthesimulations.TheinputparametersandvariablesareshowninTable7.1.ThesurfacediffusioncoefficientDswassetatzero,assumingirreversibleadsorptionoftheproteins.Forthesamereason,thedesorptionofproteinskdissetzero.Anestimateofthemaximumsurfacecapacitywasmadebydividingthenumberofmolesperprotein(1/NA)bythediameteroftheproteinsquared(dp2=4rp2).TheradiusofproteinAwasvariedduringthesimulation.One of the values is 2 nm,which is approximately equal to themeasured height ofproteinA adsorbedona surface [5, 6] andone value is 5 nmwhich is equal to theStokes radius of protein A [2]. For steady, laminar flow between two fixed parallelplates,themaximumvelocityisgivenby3/2vmean[7],wherevmeanisequaltotheflowrateQ dividedby the cross‐sectional area,which is givenby theheighth times thewidthwoftheflowchamber.

7.2.4SimulationresultsanddiscussionTheproteinAsurfaceconcentration,theproteinAbulkconcentrationandtheglucosebulk concentration are analysed over time. The parameters that were put into thesimulation canbe found in the section7.2.3 inTable 7.1. The variableska and max

werevaried,where max wasvariedbychangingrp inthemaximumsurfacecapacity

formula (seeTable 7.1). The cg, cpand are shown in Fig. 7.5 andFig. 7.6 for thedifferentadsorptionratesanddifferentmaximumsurfacecapacitiesrespectively.


Table7.1:SimulationparametersandvariablesParametervariables

Description Expression Value

h flowchamberheight 100μml flowchamberlength 6mmw flowchamberwidth 2mmlc sensingwindowlength 4mmwc sensingwindowwidth 100μmDg Diffusioncoefficientglucose 6.7x10‐10m2/s[1]Dp Diffusioncoefficientproteins 4.825x10‐11m2/s(basedon

Stokesradiusof5nmandStokes‐Einstein‐Sutherland

equation6

b

stokes

k TD

r ,with

kbtheBoltzmannconstant1.3806488x10‐23J/K,T=20ºC,andtheviscosityofwaterη=8.90x10‐4Pa·s.

Ds Surfacediffusioncoefficient 0

max Maximumsurfacecapacity2

14 A pN r

variedinsimulation:(69,17,4.4,1.9,1.1,0.69,0.17)x103fg/mm2

basedonvaluesofproteinAradiusrpof0.5,1,2,3,4,5,10nm(ProteinAStokes’radius=5nm[2])

vmax Maximumvelocity 32 meanv [7]

32Qhw

1.25cm/s

ka AdsorptionrateconstantofproteinA(adsorptionrateconstantofglucoseissetatzero)

variedinsimulation:0.1,1,10,50,100,200,400,600,800,1000,2000m3/mol/s

kd Desorptionrateconstant 0Q Flowrate 100μl/minMp MolecularweightproteinA 42kDa[2]=42x103g/molNA Avogadroconstant 6.02214x10231/molc0,g Startconcentrationofglucose 0.616%(weight

percentage)=6.16mg/ml=34.193mol/m3

c0,p Startconcentrationofproteins 40μg/ml=9.52x10‐4mol/m3

142 Chapter7

Fig. 7.5: a) Development of Cp over time for different adsorption rates of protein A, b) thedevelopmentof overtimefordifferentadsorptionratesofproteinA,c)thedevelopmentofCgovertimeasafunctionoftheadsorptionratesofproteinAandd)theCpand forka=105m3mol‐1s‐1 illustrating the time delay of the bulk concentration compared to the surfacecoverage.

Fig.7.5showsthattheproteinAbulkconcentrationstartsincreasingveryquickly.

The proteins arrive quickly in the centre of the flow chamber because of the highvelocity in the centre of the flow chamber. However due to the laminar flow, thevelocityatthesurfaceiszero.Thismeansthatintheregionthatweconsiderasbulk(≈200nmabovethesurface)thevelocityisnearlyzeroandthatdiffusiontransportstheproteinsinthisbulkregion.Thediffusioncoefficientofglucoseinwaterishigherthanproteinsinwater[1,2].Therefore,itislikelythatthediffusionofglucoseinPBSisalsohigherthanthediffusionofproteinsinPBS.Asaresult,thebulkconcentration

0 5 10 15 20 25 30

0

10

20

30

40

0 5 10 15 20 25 300

2

4

6

8

0 5 10 15 20 25 30

0

2

4

6

8

0 5 10 15 20 25 30

0

1

2

3

4

5

ka=100m3mol‐1s‐1






C p(g/m

l)

Time(s)

x102

(fg/mm

2 )Time(s)

ka=100

ka=101

ka=102

ka=103

ka=104

ka=105

m3mol‐1s‐1

c d

b







C g(mg/ml)

Time(s)

a

0

2

4

6ProteinA,ka=10

0m3mol‐1s‐1

C p(g/m

l)

x101x101

C p(g/m

l)

Time(s)

Bulk

ka=105m3mol‐1s‐1 x102

0

2

4

6

8 Surface

(fg/mm

2 )


Fig.7.6: a) TheCp over timeas a functionof max determinedby changing the radius of the

proteinA,b)azoominofthefirst30secondsofa),c)thecorresponding andd)theCgover

timeasafunctionoftheproteinA max .

of the glucose should increase earlier than the bulk concentration of the protein A,which is also found as a result from the simulations as shown in Fig. 7.5c. Theadsorption rate of glucose is set at zero aswe expect it not to bind to the sensingwindow.FromthesimulationsseeninFig.7.5c,wecanconcludethattheglucosebulkconcentration is almost not affected by the ka of protein A. On the other hand, theprotein A bulk concentration as well as the protein A surface concentration dostronglydependon theadsorption rateka (seeFig.7.5a).Asexpected, forhigherkavalues, the slope of the surface concentration is higher (see Fig. 7.5b), because theproteinshaveahigheraffinitytobindtothesurface.Moreover,forhighkavalues,the

0 100 200 300 400

0

10

20

30

40

0 5 10 15 20 25 30

0

10

20

30

40

0 100 200 300 4000

1

2

3

4

5

6

7

8

0 5 10 15 20 25 30

0

2

4

6

8

c

b

C p(g/m

l)

Time(s)

max=6.9x104

max=1.7x104

max=4.4x103

max=1.9x103

max=1.1x103

max=6.9x102

max=1.7x102

[fg/mm2]

C p(g/m

l)Time(s)

max=4.4x103

max=1.9x103

max=1.1x103

max=6.9x102

max=1.7x102

[fg/mm2]

max=6.9x104

max=1.7x104

max=4.4x103

max=1.9x103

max=1.1x103

max=6.9x102

max=1.7x102

[fg/mm2]

dx104

(fg/mm

2 )

Time(s)

max=6.9x104

max=1.7x104

max=4.4x103

max=1.9x103

max=1.1x103

max=6.9x102

max=1.7x102

[fg/mm2]

C g(mg/ml)

Time(s)

a

144 Chapter7

proteinAbulkconcentrationgetsatimedelaycomparedtothesurfaceconcentration(see Fig. 7.5d). Due to the strong adsorption rate, the proteins bind to the surface.Whenthefirstproteinsreachthesurface,theproteinAsurfaceconcentrationstartstoincrease, which continues until the maximum surface capacity is reached. As aconsequence,noproteinscanbindtothesurfaceanymore.Thebulkregion(≈200nmabove the surface) gets depleted for the time that the proteins can still bind to thesensorsurface.Aftersaturationofthesurface,theproteinAbulkconcentrationstartsincreasing as it equilibrates with the centre region of the flow chamber in whichconstantlyproteinAwithconcentrationof4μg/mlarrives.

Forka=105m3mol‐1s‐1,thesimulationswererepeated,butfordifferentmaximumsurfacecapacities max (seeTable7.1).Fig.7.6a/bshowthatforahigher max ,ittakes

moretimebeforetheproteinAbulkconcentrationincreases.Fig.7.6cshowthatthereisno timingdifference for theproteinA surface concentrationasexpected,becausetheadsorption ratewasconstant for this setof results.Consequently, ahigher max

results in larger timedelaysofproteinAbulk concentration compared toproteinAsurfaceconcentration.Furthermore,itcanbenoticedthattheendlevelofthesurfaceconcentrationisequaltothemaximumsurfacecapacity max .Again,theglucosebulk

concentrationisalmostnotaffectedbytheproteinAbindingatthesurfaceasshowninFig.7.6d.

Toconclude,wecansaythatthesimulationresultscanexplainthetimedelayoftheproteinAbulkcomparedtotheproteinAsurfaceconcentration,aswasfoundinthe analysis of the measurements where only protein A was added to the buffersolution in high concentration (see Fig. 7.1). This delay strongly depends on theadsorptionratekaandthemaximumsurfacecapacity max .Forahighka,allproteins

which reach the surface bind to the surface, which causes a depletion of the bulkregion(≈200nmregionabovesurface)until thesurfaceis fullycovered(maximumsurfacecapacity max isreached).Ontheotherhand,thetimedelayoftheD‐glucose

bulksignalcomparedtotheproteinAbindingsignalisadifferentstory.Asexpectedthe D‐glucose signal always starts earlier than the protein A signal, because of thehigher diffusion coefficient of D‐glucose, and is not affected by the protein A(independentofkaand max ofproteinA).

7.3TimedelayBSAandD‐glucose

To verify that the time delay of bulk signal of D‐glucose compared to binding ofproteinA isnotproteinAspecific,D‐glucosewasalsoaddedsimultaneouslyto66.4kDaA7906‐100g bovine serumalbumin (BSA) in a 1x PBS buffer. This proteinwas


Fig.7.7:a,b)Twoexperimentswhereshowing effNovertimeat457,561and660nmadding

proteinAandD‐glucosesimultaneouslyto thewaveguide,c,d) thecorresponding n duetothebulk and thebindingof theproteinsdeterminedwith the combinedapproachande, f) azoom‐inofc,d).

0 500 1000 1500

0

1

2

3

4

0 500 1000 1500 2000 2500

0

2

4

6

0 500 1000 1500

0

2

4

6

8

10

12

0 500 1000 1500 2000 2500

0

1

2

3

4

300 400 500 600 700‐4

‐2

0

2

4

6

8

1200 1300 1400 1500 16000

2

4

6

8

10

12

Neff

Time(s)

660nm561nm457nm

x10‐4

2g/mlBSA+6.16mg/mlD‐glucose

PBS

PBS

Neff

Time(s)

660nm561nm457nm

x10‐4

6.16mg/mlD‐glucose+100g/mlBSA

PBS


PBS

PBS

n p

roteinA

Time(s)


0

2

4

6

8

10

12

bulk

n D

‐glucose


n p

roteinA

Time(s)


0

2

4

6

8

10

12

bulk

n D

‐glucose

x10‐2 x10‐4dcore=66.5nm

x10‐2 dcore=66.5nm x10‐4

n p

roteinA

Time(s)

BindingBSA(2nm)

‐1

0

1

2

3

4

bulk

n D

‐glucose


n p

roteinA

Time(s)

BindingBSA(5nm)

‐1

0

1

2

3

4

c

ba

bulk

n D

‐glucose

fe

d

146 Chapter7

dissolved in a salt bufferwhichwas replaced by the 1x PBS buffer using a ZebaTMDesaltSpinColumn.Becausesomelossofproteincanbeexpectedafterusingthespincolumn, the concentrationwas determined afterwards by aNanoDrop®UV‐VisND‐1000 Spectrophotometer. Results of two experiments where D‐glucose was alsoadded simultaneously to two different concentrations of BSA are shown in Fig. 7.7.Alsoforthismeasurementitisdifficulttodeterminetheexacttimedelaybecauseofthefeaturesinthemeasurementsduetocuttingoftheinterferencepattern.However,it can be concluded that there is a time delay of the bulk signal compared to thebindingsignalduetoBSAwhichisconcentrationdependent.ThismeansthatthetimedelayisnotonlyobservedforproteinA.

7.4Timedelayasfunctionoftubelength

As the simulations confirm that time delay of the bulk signal due to D‐glucosecomparedtothebindingofproteinsisnotcausedintheflowchamber,thetimedelayshouldbecausedbeforethemoleculesreachtheflowchamber.Tocheck if thetimedelayisinfluencedbythetubeswhichguidethebuffercontainingthemoleculestotheflow chamber, the tube length was varied. D‐glucose (6.16 m/ml) and protein A(1 μg/ml) were added simultaneously to the different flow chambers which wereconnectedtotubeswithdifferentlengths.Toenlargethetubelength,extratubesof40cmwereinterposedbetweenthetwonormallyusedtubesof40cmeach.Theresultsof themeasurements are shown in Fig. 7.8. Themeasurement of Fig. 7.8j does notshowthe effN ofthe660nmlaser,becausethissignalwasverynoisyandunusable,

butitcouldbeanalysedbasedonthemeasured effN ’sat457and561nm.

The timedelaywasdeterminedas the timedifferenceof thestartof increaseofthe n bulksignalandthestartofincreaseofthe n signalduetothebinding.ThetimedelayisplottedasafunctionoftheextraaddedtubelengthinFig.7.9.Theerrorbarsarequitelarge,becauseitisdifficulttodeterminetheexacttimepointwhenthesignalstartstoincrease.Despitethelargeerrorbarsit isclearlyobservablethatthetimedelay increaseswith increasing tube length.Therefore, it is likely that thetimedelayiscausedinthetubing,beforethemoleculesreachthesensingwindow.


Fig. 7.8: The effN over time where first D‐glucose was added separately and later

simultaneouslywithproteinAtothesensorforanextratubelengthofa)0cm,d)40cm,g)80cmandj)160cmandb,e,h,k)thecorrespondingdetermined n forthebulkandtheproteinAbindingusingthecombinedapproachandc,f,i,l)azoom‐inofb,e,h,k).

0 1000 2000

0

1

2

3

4

5

0 1000 2000

0

1

2

3

4

130014001500160017004

6

8

10

12

14

16

0 1000 2000

0

1

2

3

4

5

0 1000 2000

0

1

2

3

4

11001200130014001500‐4

‐2

0

2

4

6

8

0 500 100015002000

‐1

0

1

2

3

4

0 500 100015002000

0

1

2

3

4

11001200130014001500‐8

‐6

‐4

‐2

0

2

4

0 500 100015002000

‐1

0

1

2

3

4

0 500 100015002000

0

1

2

3

4

900 1000110012001300‐4

‐2

0

2

4

6

x10‐4

Neff

Time(s)

660nm561nm457nm


PBS


PBS

PBS

nproteinA

Time(s)


0

2

4

6

8

10

12bulk

nD‐glucose


nproteinA

Time(s)

BindingproteinA(2nm) ‐1

0

1

2

3

4

5

bulk

nD‐glucose


x10‐4

Neff

Time(s)

660nm561nm457nm

PBS


PBS


PBS

nproteinA

Time(s)


0

2

4

6

8

10

12bulk

nD‐glucose

x10‐2 dcore=65nm x10‐4

nproteinA

Time(s)


‐1

0

1

2

3

4

5

bulk

nD‐glucose


Neff

Time(s)

660nm561nm457nm

x10‐4

nproteinA

Time(s)



0

2

4

6

8

10

12bulk

nD‐glucose

nproteinA

Time(s)



‐1

0

1

2

3

4

bulk

nD‐glucose

Neff

Time(s)

561nm457nm


PBS


PBS

x10‐4

PBS


PBS

PBS

PBS


nproteinA

Time(s)



0

2

4

6

8

10

12bulk

nD‐glucose

cb

nproteinA

Time(s)



a

d e f

g h i

j k l

‐1

0

1

2

3

4

bulk

nD‐glucose

148 Chapter7

Fig.7.9:TimedelayofD‐glucosebulksignalcomparedtoproteinAbindingsignalasafunctionof theextra lengthof thetubeguidingthebufferandmolecules to thesensingwindowof thewaveguide.

7.5Conclusionsandforwardlook

From the simulations we can conclude that the evolution of the bulk and surfaceconcentration of protein A strongly depends on the adsorption rate ka and themaximumsurfacecapacity max .Forakaof105m3/mol/sanda max between4.4x10‐3

and1.7x104 fg/mm2 (corresponding toaproteinA radiusof1and2nm(seeTable7.1)respectivelywhichagreeswelltothemeasuredthicknessof2.1nm(seeChapter4)),atimedelayofapproximately20sto100sisfound.ThisiscomparablewiththetimedelaybetweenthebulkandbindingsignalofproteinAdeterminedwiththesize‐selective detection analysis approach. As the time delay can be explained by thesimulationswhichshowa timedelaybetweenof thebulksignaldueto theproteinscompared to the binding of the proteins, the time delay determined with the size‐selectivedetectionappearsrealandnotanartefactoftheanalysisapproach.

To verify the results of the simulations, protein A and D‐glucose were addedsimultaneously to the sensor. Simulations show that due to the high diffusion ofD‐glucosethissignalshouldbeearlierthanthebindingofproteinA, independentofkaand max . However, again the determined signal of n due to binding of proteinA

occurred earlier than the n due to D‐glucose. Therefore, this time delaymust becaused before the protein A and D‐glucose reach the flow chamber. This wasconfirmedby an experimentwhich showed that the timedelay is influencedby thetubelength.Longertubelengthsresultinalongertimedelay.ThedelayofD‐glucosecomparedtobindingofproteinAalsooccurswhenaddingtheD‐glucoseandproteinAseparatelytodifferenttubesguidingthemoleculestodifferentflowchamberwith

‐20 0 20 40 60 80 100 120 140 160 1800

40

80

120

Timedelay(s)

Extratubelength(cm)


sensingwindows.Therefore,thetimedelaycannotbeexplainedbyphaseseparationof the two different molecules in the tubes, which can occur for example for twoincompatiblepolymersincapillarytubes[8].Thiswasalsoverifiedinanexperimentwhere the same protein A concentration (80 μg/ml) and D‐glucose concentration(6.16mg/ml)wereaddedinamountof≈3.5mltoa5ml,12mmwidetransparenttube. After mixing the molecules and waiting for a day no RI differences wereobservedbetweendifferentareas.AlsoafteraddingaCoomassieBrilliantBlueG‐250dye (is used forBradford protein assays),which changes froma brown colour to amore blue colour when binding to a protein, no clear colour differences wereobservedbetweenpossible areaswithD‐glucose andpossible areaswithproteinA.Consequently, there is no strong phase separation of protein A and D‐glucose,meaning that this cannot explain the time delay of the D‐glucose compared to theproteinAmeasuredwiththeYIsensorincombinationwithsize‐selectivedetection.

Alternatively, there might be a mechanism which can transport the protein Afasterthroughthetubes,possiblyalongthewallsofthetubes.ThismechanismisnotproteinAspecific,asthedelaywasalsoobservedusingBSAinsteadofproteinA.Moreresearchisrequiredtostudytheeffectofthetubesandtheirwallsonthetimedelay.Tovisualizethedifferenceinthetubesitmightbepossibletolabel(dye)theproteinsand D‐glucose. It is also possible to determine the influence of the tubing by usingdifferent tubing materials such as (PEEK) Polyetheretherketone instead of Tygontubing.However,inviewoftime,theseexperimentswerenotdoneandthisresearchcouldnotbecontinued.

Acknowledgements

IwouldliketothankDr.ChristianBlumandDr.SaskiaLindhoudfortheirinput,ideasanddiscussionsabout timingdifferencesbetween theproteinAandD‐glucose.Thiswork is supported by NanoNextNL, amicro and nanotechnology consortium of theGovernmentoftheNetherlandsand130partners.

References

1. D. R. Lide, CRCHandbook of Chemistry and Physics, 84th Edition (Taylor & Francis,2003).

2. I.Björk,B.‐Å.Petersson,andJ.Sjöquist,“SomePhysicochemicalPropertiesofProteinAfromStaphylococcusaureus,”Eur.J.Biochem.29,579‐584(1972).

3. R.Hansen,H.Bruus,T.H.Callisen,andO.Hassager,“TransientConvection,Diffusion,andAdsorptioninSurface‐BasedBiosensors,”Langmuir28,7557‐7563(2012).

4. T. Gervais and K. F. Jensen, “Mass transport and surface reactions in microfluidicsystems,”Chem.Eng.Sci.61,1102‐1121(2006).

150 Chapter7

5. M. C. Coen, R. Lehmann, P. Groning, M. Bielmann, C. Galli, and L. Schlapbach,“AdsorptionandbioactivityofproteinAonsiliconsurfacesstudiedbyAFMandXPS,”J.ColloidInterfaceSci.233,180‐189(2001).

6. S. Ohnishi, M. Murata, and M. Hato, “Correlation between Surface Morphology andSurfaceForcesofProteinAAdsorbedonMica,”Biophys.J.74,455‐465(1998).

7. B.R.Munson,D.F.Young,andT.H.Okiishi,“Fundamentalsoffluidmechanics,”5thed.(Wiley,2006),p.323.

8. R. H. Tromp and S. Lindhoud, “Arrested segregative phase separation in capillarytubes,”Phys.Rev.E74(2006).

Chapter8

Applicationsforsize‐selectivedetection:aforwardlook

AbstractThis chapter explores what are the most relevant applications for the Younginterferometer (or a similar) sensor in combination with size‐selective detectionbasedonmultiplewavelengths. Inthischapterwealsoaddresswhatshouldbenextstepsinfutureresearchbasedontheconclusionsdrawnfromtheotherchapters.Forthat purpose, a small technology assessment study was performed and a mini‐workshop was organised to identify possible innovation pathways and relevantfactorstoconsiderfromatechnological,productandcommercializationpointofview.From the assessment it was determined that the added value of the sensor is tomeasure size‐selectively (discriminate between substances based on their size) andthatthegaininsize‐selectivityshouldbeworththelossinsensitivityandstabilityandthe increasing costs. Therefore, pharmacy and themedical sector are probably lesssuitablemarkets for this sensor, as in these application areas, specific biochemistryand sensitivity are more important than size‐selectivity. A research platform foruniversitiesorcompanieswasfoundtobethemostpromisingmarket,butitmustbenoticedthatmarketsshouldbeanalysedinmoredepthforamorecompletepicture.Possibleapplicationsinthismarketaremeasurementtoolstomeasurethekineticsoflayer growth, a signal enhancer, a quick screener or an alternative for a ScanningElectron Microscope, an alternative for Dynamic Light Scattering (to measurethicknessoflayersorparticles)andaresearchtooltostudyfundamentalresearchonbindingorothersurfaceeffectsorflowdynamics.Thisresultedinresearchquestions“Whataretheminimumandmaximumsizewhichcanbedetectedanddetermined?”and“Whatdifferencesinsizescouldbediscriminatedfromeachother?”.Furtherresearchisrequired to answer these questions. When a certain application or market isidentified,thenthesensorshouldbeoptimizedtodeterminethelimitsofthedeviceforthespecificapplicationinmind.

152| Chapter8

8.1Introduction

Integratedoptical(IO)sensorshavebeendemonstratedasapowerfuldetectiontoolfor biosensing, because of their ability to detect label‐free, in real‐time and verysensitive.TheYounginterferometer(YI)isaverypromisingbiosensorbecauseofitscapabilityofmultiplexinganditsextremelyhighrefractiveindex(RI)sensitivity(10‐7‐ 10‐8 RIU) [1, 2]. However, non‐specific signals like bulk changes and non‐specificbinding still hamper a full utilization of the high sensitivity and therefore also asuccessfulapplicationof these typeofbiosensors.Wehaveextensively studiedhowthe use ofmultiplewavelengths can contribute to an increased specificity (definedhereastheabilitytocorrectlydetectanalytes,meaningnotmeasuringfalsepositives).Itwas foundthat thedifferentevanescent fieldsof thedifferentwavelengthscanbeused to differentiate between particles based on their size and therefore add size‐selectivity to the sensor. Experimental data showed that 85 nm beads could bediscriminatedfromproteinA,whichisfoundtobe≈2nminheightbyotherstudies[3,4],andD‐glucoseinducingahomogeneousRIchangeinthewholeevanescentfield(few hundred nanometres). Therefore, for specific cases, it is possible with size‐selectivedetectiontodiscriminatebetweenspecificbindingoflargermolecules,non‐specificbindingofsmallermolecules,and/orbulkeffects,toincreasethespecificityofthe sensor. However, it should be explored what will happen if the size of thesubstancesisnot40timeslargerbutonly2or3timeslarger.Nevertheless,wewerealsoabletoobservetimingdifferencesbetweentheproteinAandtheD‐glucosewhichwasnotpossiblebeforewhenmeasuringwithasinglewavelength.

Moreover, we presented a theoretical approach to discriminate between RIchanges from different layers in the evanescent field based onmeasurementswithdifferent wavelengths. This method requires input of theoretical waveguideparameterswhichdonotalwaysfitwiththemeasurements.Somemeasurementswithsimilar beads did not fit the theoretical model andwere different when theyweremeasuredindifferentsolutions.Therefore,amuchmorepragmaticwavelengthratio‐based approachwas developedwhich only requires calibration experiments of thesubstancesalone.ThismethoddoesnotrequireanyinputoftheoreticalsizesandRI’sof the waveguide and the measured substances, but is just based on the response(ratiosof effN ’softhedifferentwavelengths)ofthesensortothesubstanceswhich

are measured. Information on the absolute amplitude of n is lost using thisapproach,however calibrationmeasurementswhichareusually also requiredusingan YI with a single wavelength can be performed to determine an analyteconcentrations.D‐glucose,85nmbeadsandproteinAwerediscriminatedfromeachother using this ratio‐based approach. The combined theoretical and ratio‐basedapproachwasdeveloped to determine the absolute value of n , the correspondingsurfacemasscoveragesand/orthesizeofthemeasuredsubstances.

Applicationsforsize‐selectivedetection:aforwardlook |153

Calculationsshowthatbyimprovingthespecificityintheformofsize‐selectivity,thesensitivityisreduced.Moreover,experimentsandsimulationsshowthatthemoredifferentsources(eachinducinga n )youwanttodiscriminatefromeachother,themoreallnoisesourcesareenhancedandemergeinthedata.Atthemomentthesetupis limited to use twowavelengths to reliably discriminate between two substances.When discriminating between three substances is required, the setup needs to beimproved. Especially artefacts due to insufficiently reduced boundary effects of theinterferencepatternshouldberemovedfromthemeasuredphasechanges.Measuringwithmultiplewavelengthsandpolarizations,wetriedtoimprovethesetup.Becauseof themoredifferent responseof certainwavelength/polarization combinations thelossinsensitivitycanbereduced.Thiswasrealisedfordiscriminationbetweenthreesubstances, however, the aforementioned artefacts in the measurement were alsoincreasedbythesemeasurementsmeaningthewewerenotabletoimprovethesetupin thisway. Further research is required to realise optimal detectionwithmultiplewavelengthsandpolarisations.

As the sensitivity is reducedwhen discriminating betweenmultiple substances,the techniqueofsize‐selectivedetectionmightbe lesssuitable for theapplicationofbiosensing,asgenerallyhighsensitivitiesarerequiredtodetectlowconcentrationsofbiologicallyrelevantanalytes.Toexplorewhatarethemostrelevantapplicationsforthe YI (or a similar sensor) in combination with size‐selective detection based onmultiple wavelengths and what should be next steps in future research, a smalltechnology assessment (TA) studywas performed. Aminiworkshopwas organisedtogetherwithDr.HaicoteKulveandDr.VerenaStimberg,bothpostdoctoralfellowsinthedepartmentofScience,TechnologyandPolicyStudies(STePS)attheUniversityofTwente. A general introduction to TA, themethod (workshop) and the results andfindings of this workshop are presented in section 8.2. Section 8.3 presents thereflectionsonthisworkshopandthethesisingeneralandpresentsaforwardlook.

8.2Technologyassessment

ThemaingoaloftheTAmini‐workshopwastoobtaininputforfurtherresearchandto look for possible applications and/or innovation pathways. This means that wewanted to know what are interesting applications for the sensor as we look at itscapabilities and what questions needed to be answered for that and what futureresearchneedstobeperformed.Inthissectionwegiveasmall introductionintoTAandweexplainwhywechoseforamini‐workshopandwhatpreparationsweretakenforthisworkshop.Afterthatwereportthemostimportantfindingsoftheworkshopandwepresent conclusions fromtheTAexercise togetherwith the implications forfurtherresearch.

154| Chapter8

8.2.1IntroductionTo further the development and understanding of the YI in combinationwith size‐selectivedetection,wecanandwillbuildupontheinsightsgainedduringthisproject.However,what is interesting froma scientific point of viewmaynot necessarily beimportantfromavalorisationorknowledgeutilizationperspective.WorkingwiththeDutchNanoNextNL programwewere inspired by its ambition to not only performcutting‐edge fundamental and technology research but also to innovate and bringmicro‐andnanotechnology‐enabledtechnologiesclosertothemarket.That is,whatcould be possible applications of the YIwith its size‐selectivity possibilities andwhatwould this imply in terms of further research questions? These questions open‐up arangeof considerations for the introductionofnew technologieson themarket (i.e.their introduction and acceptance by firms, regulatory authorities and the generalpublic). Thus, another set of evaluation criteria and perspectives come into playcompared with a researcher’s purely scientific perspective. Examining andanticipating the perspectives of other actors involved in the introduction anddevelopment of new technologies such as the YI with size‐selective detection thenhelpstoenrichaviewonwhataresocietallyrelevantissuesassociatedwiththeYIaswell as related research questions. Taking into account these issues and questionsmaysupportaccelerationofcommercializationoftheYIwithsize‐selectivedetectioncapabilities. The perspectives of a broader set of actors may help to open up newresearchavenueswhichhavenotyetbeenconsideredbefore.

TobroadentheperspectiveonpossibleroutestofurtherresearchanddeveloptheYI with size‐selective detection capabilities, we make use of techniques andapproachesfromthefieldofTA.TAinvolves“theearlyidentificationandassessmentofeventualimpactsoftechnologicalchangeandapplications,performedasaserviceto policy making and decision making more generally” [5]. There exists a broadvariety of approaches in this field. Here, we make use of a particular perspective,namely constructive technology assessment (CTA) [6] and how it is applied anddeveloped for emerging technologies such as nanotechnologies [7]. WhatdistinguishesCTAfromotherTAforms is that ithasastrongfocusontechnologicaldevelopments and their dynamics and aims to feedback insights of assessments toongoingtechnologydevelopments.Asweareinterestedinenrichingourperspectiveoninterestingresearchquestionswithconsiderationsofsocietalaspects,CTAseemswell suited forourpurposes. In thenext sectionwewill elaborate theapproachwefollowedtoapplyaCTAperspectivetothisproject.

8.2.2MethodToexplorepossibleapplicationsoftheYIwithsize‐selectivedetection,weorganizedasmallinteractiveworkshop.Tosupportdiscussionsduringtheworkshopweappliedaso called, ‘multi path mapping’ tool [8]. This tool shares similarities with what is


knownasroadmapping.Roadmappingisawell‐knowntechniquewithincompaniesaspartoftheirlongtermplanningstrategies[9].AfamousexampleistheInternationalTechnologyRoadmapforSemiconductors.Specificforthemultipathmappingtoolisthat it (1) identifies different routes, ‘paths’, towards the development andintroduction of products and (2) that it starts from ongoing research activities andtakes into account the inherent uncertainty and open‐ended character of scientificwork. The mapping starts from identifying research lines or functionalities oftechnologies,thentheirintegrationintoaplatformorproduct,followedbyaspecificapplication area (a product/market combination), and finally the consideration ofbroader societal aspects regarding the specific application. Mapping potentialapplicationsfromresearchlinesorenvisionedfunctionalitiesofspecifictechnologiesthenresultsinavarietyofpossibleroutestowardsapplications.Thefunctionofthistoolisthreefold.Firstly,itsupportsaresearcher,orinthiscaseagroupofworkshopparticipants,toarticulatethesedifferentstepsandconsiderpuzzlesandchallengesineachstep.Secondly,itsupportsthemtothinkaboutpossibleapproachestodealwiththesepuzzles and challenges andoffers an overview.Thirdly, it gives indications ofwhat could be possible main pathways along which these envisioned applicationscould develop. As these technologies become more developed, and all kinds oftechnological and social relationships come into place, it becomes increasinglydifficult to deviate from these routes.1 Mapping such pathways then enables theresearcherorworkshopparticipantstoidentifyandbecomeawarewhereitbecomesmore difficult to deviate from such application routes (e.g. when do you get at thepointwhereitbecomesreallydifficulttochangegears,becauseofsunkinvestments?).

Constructingamulti‐pathmapcanbedonebyoneself,butitbecomesmorerobustwhenmorepeopleandorganizationsare involvedwithdifferentperspectivestothetechnology. Ideallyallstakeholdersrelatedto, inourcasetheYI,wouldbe involved.However,practicallyandbecauseoftheuncertaintiesinvolvedwithnewtechnologiesandtherefore thechallenges toassess futureapplications, thismaybedifficult.Thiswasthecaseforthisproject,wheretherewasastrongfocusonfundamentalresearchwith many open‐ended questions. Furthermore, an expansive exploration ofapplicationswasbeyondthescopeofthisresearchproject.Wethuschosetofocusinparticular on the lower layers of the multi path mapping tool. That is, assessingtechnological building blocks, possible functionalities and application areas.Therefore, we opted for inviting a mix of researchers with varying disciplinarybackgroundsandresearchorientations,andcompaniesinterestedintheYIwithsize‐selectivedetection.Thisapproachlookedpromisingandproductivewithinthescopeof thisproject inorder to take intoaccountabroadersetconsiderations for furtherresearch.

1Thisphenomenonisknownas‘pathdependency’and‘pathcreation’.

156| Chapter8

Wedesignedaworkshopbasedonamulti‐pathmapinwhichwepresentedafirstroundassessmentofdifferentroutestowardspossibledifferentapplicationsbasedonresearch findings. During the workshop the participants of the workshop wereencouraged to change themulti‐pathmapandadd inputs to themulti‐pathmapbyplacing post‐it notes on the multi‐path map (see Appendix 8.B). Furthermore, wemadeaworkshopprogramincludingquestionspersessionwhichcouldbeaskedtotheparticipantstoencouragethediscussion.Themulti‐pathmap,workshopprogramand the workshop participants are presented in Appendix 8.A. The goals of theworkshopwere:

I. Joint development of amulti‐pathmap for commercialisation of the Younginterferometerincombinationwithmultiplewavelengths.

II. Identification of factors to consider in the multi‐path map, including a)technological/product development issues, b) commercialization aspects,including user/industry requirements; c) possible broader societal aspectssuch as (health, environmental and safety) risks, regulations (standards,rules/regulationsinparticularapplicationdomains),publicacceptance.

III. Implications of the assessment in terms of (most) plausible and promisinginnovationpathways,importantnextstepsforfurtherresearch.

Considering the focusof theworkshopandthe limitedamountof time,prioritywasgiven to points i, ii a‐b. Theworkshopparticipantswere invitedby sending themageneral letter including the goals of theworkshop, theworkshop program and themulti‐path map. Furthermore, a presentation was prepared including backgroundinformation on the YI in combination with the size‐selectivity based on multiplewavelengths and the most important experimental and computational results, toinformworkshopparticipantsaboutthisresearchproject.

8.2.3FindingsworkshopIn this section we discuss the outcomes of the workshop. The findings from theworkshoparepresentedinTable8.1ofwhichwediscussthemostimportantfindings,basedontheirimpactonpossibleapplications,here.

Acrucialpointduringtheworkshopdiscussionwasthemeaningofspecificityandsize‐selectivity, because there were different interpretations of the meaning ofspecificitywhichcould leadtodifferent interestingapplications. Inbiochemistrythespecificity relates to the selective attachment or influence of one substance onanother, for example an antibody and its specific antigen, but in general it can alsomean the ability to correctly detect specific analytes, meaning not measuring falsepositives.Thesize‐selectivemeasurementscan improve thespecificityof thesensorusingthesecondmoregeneraldefinition.However,usingthefirstdefinitionfromthebiochemistryperspective,thisisnotpossibleassize‐selectivitydoesnotimprovethe


selective attachment of substances2. Therefore, itwasdecided tonot talk about thespecificityofthesensorduringtheworkshop,butusingthetermsize‐selectivitywhentalking about discrimination between substances based on their size using themultiplewavelengthsincombinationwiththeYI.Furthermore,itwasmentionedthattheYIusingasinglewavelengthisgenerallyafactoroftentimesmoresensitivethanSPR, but the SPR is a more established technique. The advantage of the YI withmultiple wavelengths is the ability tomeasure size‐selectively. However, there is atrade‐off:thegainintermsofsize‐selectivityresultsinenhancementofnoiseanddriftand therefore in a less stable signal and a reduced sensitivity (comparable to SPRwhenusingtwowavelengthstodiscriminatebetweentwosubstances).Therefore,itisimportanttofindapplicationswhichrequiresize‐selectivityandfastdetection(whichis still possiblemeasuringwithmultiplewavelengths) andwhere sensitivity is lessimportant.Thereby,itisimportanttonoticethatdiscriminatingbetweenacontinuousdistributionofsize isnotpossible,so it is importanttosearchforapplicationsweretwo(or three,whensetup is improved)particleswithmeasurabledifferent sizesorlocationshavetobediscriminated.Alternatively, thesize‐selectiveYIcanbeusedtodeterminethesizeofcertainsubstances.

Withall these issuesandpossibilities in termsofmeasurementtime,sensitivity,specificity and size‐selectivity, this led to a lot of suggestions of participants forpotential applications for the YI sensor with size‐selective detection. A possibleapplication is thedeviationbased sensor forwaterqualitywhere theyuseRI as anindicatorofthewaterquality[10].Size‐selectivemeasurementsmightbeinteresting,assurfaceandbulkeffectscanbeseparated.Onecanalsothinkofapplicationswheresize of particles is very important, for example in air pollution, because smallerparticlesaremorehazardous than largerparticles. It canalsobeused fordetectionand/orcharacterisationofnanoparticlesusedinrubbertirestoreducetheirabrasion.Otherwise, itmight be possible to detect or characterise nanoparticles in seawater,asbestos or new materials in material science. Furthermore, ascertaining theauthenticityofproductsinapharmaceuticalplantorafoodcompanywassuggestedasanapplication.Thesize‐selectivedetectioncouldbeusedtodetermineifproductsareindeedtheproductswhichwereexpectedbyverifyingthesizeoftheproductse.g.incombinationwithspecificantibodies.Anotherapplicationmightbemonitoringtheproductionofcertainchemicalsathighspeedwhichrequireaspecificsize.

The sensor might also be used as alternative for or an addition to othertechniques.Itcanbeusedtomeasurekineticsof layergrowthordeterminelayerorparticle thickness such that it can be used as an alternative for dynamic lightscattering (DLS).Alternatively, the sensormightbeusedasaquick screenerwithayes/noanswerbasedonsizebeforeamoreelaboratestudyisdonewithforexamplea

2 In general it is important to discuss in detail important characteristics of the sensor likesensitivityandspecificityduringamappingexerciseasitcanleadtodifferentoutcomes.

158| Chapter8

scanningelectronmicroscope(SEM)orevenasanalternativeofSEMastheYIwithsize‐selectivedetectionisaneasierandfastermethod.However,theSEMcanstillgiveadditionalinformationasitallowstovisualiseparticles,wheretheYIdoesnot.

Biochemistry plays an important role in biosensors, so there were somesuggestions to combine biochemistry with size‐selectivity. The sensor might beapplicableforvirus(50‐200nm)detectionwherenon‐specificbindingofproteins(1‐10nm)playsarole.Alternatively,specificantibodiescouldbeconnectedtoaspacertoseparatethemphysicallyfromnon‐specificbindingofparticlesatthesurface,suchthat a difference in size between specific analytes and non‐specific particles is notrequired. By rearranging the linker quickly, so bringing the analyte from a certaindistancetoveryclosetothesurface,thesignalgetsenhancedquicklyasthesignalofthebulkwilldecreaseandthesignalatthesurfaceincreases.Inthiscaseitisrequiredtoseparatethebulksignalfromthesignalatthesurfacewhichispossiblemeasuringwithtwodifferentwavelengths.Awashingstepcanbeusedtogetridofnon‐specificbound particles to the surface and particles in the bulk, but this is probably notrequired as the specific signal, coming from bringing the analyte bound to theantibodywithalinkerclosetothesurface,canberealisedveryfastcomparedtonon‐specificbindingofparticles.

Itmightalsobepossible touse thesensortodofundamentalbindingstudiesasthe sensor is very sensitive for changes at the surface and by using the multiplewavelengths even more information can be obtained. In a similar way it might bepossible to study flow dynamics such as laminar flow profiles. According to aparticipantthesensormightberelevantforworkonlab‐on‐a‐chipandresearchinthefieldofmicrofluidicsornanofluidics.Thesensitivityisnotsoimportanthere,butthelocationofparticlesinthechannelisveryimportant.Finally,itwassuggestedtousethesensorasathermometer,becauseitispossibletosuppresssurfaceeffectsandusebulk signals as a thermometer as the bulk RI is very sensitive for temperaturechanges.

Next to the applications, also possible markets were suggested for the YI withsize‐selectivedetection.Drugsdetectionwasmentionedwith the exampleof heroinandcocainewhicharedifficulttodistinguish.Asthedifferenceinaveragemassisonly66Da[11],itisprobablymuchsmallerthanacoupleofnanometreswhichwasfoundfor the thicknessofproteinA thathasamolecularweightof42kDa [12]. So theYIwithsize‐selectivedetectionisprobablynotabletodiscriminatebetweenthembasedontheirsize,buttheremightbeothersubstanceswhicharedistinguishablebasedontheirsize.

As there are a couple of applications in fundamental research, it was alsosuggested that the sensor can be used as a research platform in universities orcompaniesinwhichRIVMorTNOmightbeinterested.Inpharmacy,biochemistryisoftenmore important than size‐selectivity, so thismarket seems less suitablewhenonlytakingintoaccountthesize‐selectivity.Furthermore,medicalapplicationsseem


tobe less suitableashighsensitivity isvery important.However, these findingsarenotdefinitiveastheremightalwaysbecertainapplicationsweresize‐selectivitymightbemoreimportantthanbiochemistryorsensitivity.

The discussion in the workshop was based on the multi‐path map which ispresentedinFig.8.1,wheretheredcolouredpartswereaddedduringtheworkshop.Therewasnotalotofdiscussiononthelevelofsocietalembeddingasthefocuswasmore on the technology and functionality of the sensor and the consequentapplications,whichisprobablyaconsequenceofthebackgroundoftheparticipants.Itwas determined that the unique selling point of the sensor and its advantage overothertechniquesisthepossibilitytomeasuresize‐selectively.Furthermore,thetrade‐off between sensitivity and size‐selectivity is an important conclusion forcommercialisationbutalsoforfurtherresearch.Moreover,thediscussionduringtheworkshop resulted in a lot of new research questions which will be discussed insection 8.2.4, where also the most promising pathway in the multi‐path map ispresented.

Table8.1:Issuesandsolutionsperlevelofthemulti‐pathmapofwhichthepointsinboldarediscussedinthetext.Tablecontinuesonnextpage. Issues Solutions

Technology 1)Thetermspecificitymightbemisleading,itcanmeandifferentthings.2)Stability(forlongmeasurements)andsensitivityofthedevicearethelargestchallengesinasteptocommercialization,howeverthereisatrade‐offforsize‐selectivityandsensitivityandstability.3)Interferometerperformsgenerally10xbetterintermsofsensitivityastheSPRincaseofsinglewavelengthmeasurement.However,SPRismoreestablishedtechnique

1)Usethenameofsize‐selectivityinsteadofspecificity.2)TheaddedvalueofYIwithmultiplewavelengthscanbetheadditionofsize‐selectivity(verticalresolution)3)Findapplication(s)whichrequire(s)size‐selectivityandfastdetectionandwherehighsensitivityislessimportant

Functionality 1)Abroadparticlesizedistributionmakesitmoredifficulttodistinguishbetweenparticles.Measuringcontinuousdistributionisnotpossible.2)Whenspeakingaboutsensitivitiesoffg/mm2,beveryclearwhatassumptions(e.g.sphericalparticles,RI,coverage,etc.)aremade.

1)Searchforapplicationstodiscriminatebetweentwoorthree‘particles’withmeasurabledifferencesinsize

160| Chapter8

Issues Solutions

Product 1)Inthecaseofbiosensors,biochemistryplaysamoreimportantrolethansize‐selectivity,butthesetechniquecanbecombined.

Potentialapplications:a)Deviationbasedsensorsforwaterqualityb)Measureparticlesingas,airpollution(nanoparticles,sizeisimportant,smaller=morehazardous)c)Environmentalproblem:e.g.nanoparticlesinseawaterorasbestosd)Nanoparticlesinrubbere)Materialsciencef)Pharmacy/food:knownproductsg)Productionchemicalsathighspeedh)Kineticsoflayergrowthi)Quickscreenerj)Alternativefor/additiontoSEMk)Thicknessoflayers(alternativeforDLS)l)Virusdetectionwithalotofnon‐specificbindingm)Signalenhancern)Fundamentalresearchonbindingorothersurfaceeffectso)Flowdynamicsp)Thermometer(~bulksignal)

Market 1)Medicalapplicationsneedmostlyhighsensitivity2)Participant:startfromthetechnologyandseewhatitcando,andwhatadvantagesithasoverexistingtechnologies.Anotherparticipantarguesthatonecanalsostartfrompromisingapplicationsandreasonbacktofunctionalitiesandtechnologies.3)Inthecaseofpharmaceuticals,e.g.detectingactivepharmaceuticalingredients,biochemistryismoreimportantthansize‐selectivity

1)Issue1)and3)arenotdefinitive,becausetheremightalsobeapplicationsforwhichquickandsize‐selectivemeasurementsaremoreimportantthanbiochemistryorsensitivity,inwhichitcanbecombinedwithbiochemistry.2)Intheworldofdrugsdetection,theYIwithmultiplewavelengthsisnotprobablynotabletodiscriminatebetweenheroinandcocaine(itseemstobedifficulttodiscriminatebetweenthem)basedontheirsizeastheirmassdifferenceisaround66Da.3)Aresearchplatformwhichcouldbeusedatuniversitiesorcompaniesappearedtobethemostpromisingpathway.OrganizationssuchasTNOandRIVMmightbeinterestedaccordingtoparticipant.

Societalembedding

1)IsitaproblemiftheIPalreadyisprotectedandwoulditbepossibletocommercialisethisplatform(ingeneral)?

1)Thismightstillbepossible,butatthemomentwefocusonthetechnologyandfunctionalityandconsequentapplicationsofthesensor.Animportantquestioniswheredoesthistechnologyofferanadvantageoverexistingtechnologies.Sowhatareuniquesellingpoints?Thisistheaddedvalueofsize‐selectivity.


Fig.8.1:Multi‐pathmapafterworkshopwherechangesandadditionsmadetothemulti‐pathmapduringtheworkshoparemarkedinred.

162| Chapter8

8.2.4ConclusionsfromtechnologyassessmentandimplicationsforfurtherresearchFromthediscussionduringtheworkshopandtheissuesandsolutionsresultingfromthisdiscussionweselected,accordingtous,themostpromisingpathwayinthemulti‐pathmapasshowninFig.8.2.However,itmustbenoticedthatinnovationpathways,including impacts on society and implications for research, should be analysed inmoredepthforamorecompletepicture.Furthermore,thediscussionanditsresultingissuesandsolutionsalsoledtoresearchquestionswhicharepresentedinTable8.2.Theresearchquestionsrelatedtothemostpromisingroutearediscussedhere.

First of all, it is important to knowwhat are theminimum andmaximum sizeswhich can be detected with the devices if a size of a particle or layer should bedetermined.Whendiscriminatingbetweendifferentsizedparticlesit is importanttoknowwhatdifferencesinsizescouldbemeasured.So,thismeansthatitisrequiredtoknow if the cut‐off values differ between 1 and 10 nm or between 1 and 100 nm.Further research is required to determine these valueswhich should be comparedwiththeresolutionofalternativetechniqueslikeSEMorDLS.Intermsofsensitivity,weknowthatwe loseapproximatelyanorderofmagnitudecompared to thesinglewavelength YI when adding an extra wavelength to discriminate between twosubstancesordeterminethesizeofaparticle.However, thisshouldbe lessrelevantfortheselectedapplicationsfromTable7.1wherefastandsize‐selectivedetectionismore important than sensitivity. With fast measurements we mean that with thecurrent, far fromoptimized setup,measurements canbedonewithin5‐10minutes,but this canbeoptimized further tomeasurement times smaller thanaminute.Butmostimportant,thegaininsize‐selectivityshouldbeworththelossinsensitivityandstability,andextracosts.

Fromamarketpointofview,itisimportanttoknowwhattherequirementsareoftheparticularmarkets,iftherequirementsmatchwiththedevicespecificationsandifthis is not the casewhat could be done to reach the requirements.When a certainapplicationormarketisfound,thenthesensorshouldbeoptimizedtodeterminethelimitsofthedevice.Asmostoftheparticipantswerefocussedmoreonthetechnology,itsfunctionalitiesandtowhichproductthiscouldleadandnotoneclearapplicationwas discussed, no research questions were constructed in terms of societalembedding.


Fig.8.2: Illustrationof themostpromising routeof themulti‐pathmapbasedon the findingfromtheworkshop.

Table8.2:Researchquestionsperlevelofthemulti‐pathmap. Researchquestions

Technology 1)Cantheuseofmultiplewavelengthtoaddsize‐selectivityalsobeappliedtoothertechniques?2)Howfastcanyoumeasureyoursignal?3)Whatareapplicationswhichrequiresize‐selectivityandfastdetectionandwheresensitivityislessimportant?4)Isitpossibletotunethespatialresolutionofthedevice?

Functionality 1)Whatarethesizes(minimumandmaximum)thatcanbedetectedwiththedevice?2)Whatdifferencesinsizecanyoudetermine?Isthecut‐offvaluethedifferencebetween1nmand10nmor1nmand100nm?3)Whatisthepositionalaccuracy?4)Howsensitivecanyoumeasurewhenusingsize‐selectivedetection?5)Whatisthesensitivity/masscoverageofownexperiments?

Product 1)Whatareapplicationswhichrequiresize‐selectivityabovespecificityduetobiochemistryorsensitivity?2)Forwhichapplicationdoessize‐selectivedetectionplayanimportantroleandcoulditbeanadvantagethattheYIplatformcanbothcatchparticlesandbeabletodiscriminatethem?3)Isthegaininsize‐selectivityworththelossincosts,sensitivity,stability?4)HowcanDLSbecomparedwithYIwithmultiplewavelengthsintermsofresolution?

Market 1)Whatarerequirementsoftheparticularmarkets?2)Dotherequirementsmatchwiththedevicespecifications?3)Whatisneededtoreachtherequirementsofaspecificdevice?

Societalembedding

164| Chapter8

8.3Conclusions,reflectionsandforwardlook

ATAworkshop resulted inpromisingapplications for theYI (butalsoapplicable tootherevanescentfieldbasedsensors)combinedwithsize‐selectivedetection.Firstofall, itbecameclear that termspecificitycanmeandifferent things.Therefore, itwasdecidedtoavoidusing thetermspecificity,buttospeakaboutsize‐selectivityaswediscriminate between substance based on their size. If we compare our sensor toothertechniquestheaddedvalueofthesensoristhesize‐selectivity.However,asthesize‐selectivity negatively influences the sensitivity and stability of the sensor, wehave to look forapplicationswhich require fast and size‐selectivedetectionand forwhich sensitivity is less important. The gain in size‐selectivity should beworth thelossincosts,sensitivityandstability.Therefore,pharmacyandthemedicalsectorareprobablylesssuitablemarketsforthissensor,asusuallybiochemistryandsensitivityare more important than size‐selectivity. Different market and applications werediscussed during theworkshop, resulting in a research platform for universities orcompanies to be the most promising market, but it must be noticed that marketsshouldbeanalysedinmoredepthforamorecompletepicture.Possibleapplicationsinthismarketaremeasurementtoolstomeasurethekineticsoflayergrowth,asignalenhancer,aquickscreener forSEM,analternative forDLS(tomeasure thicknessoflayersorparticles),aresearchtooltostudyfundamentalresearchonbindingorothersurface effects or flowdynamics. This resulted in research questions “Whatare theminimum and maximum size which can be detected and determined?” and “Whatdifferences in sizes could be discriminated from each other?”. Further research isrequired to answer these question and to determine these valueswhich should becomparedwith the resolution of alternative techniques.Whenwe look at it from amarket perspective, it is important to define what the requirements are of theparticularmarkets,iftherequirementsmatchwiththedevicespecificationsandifnot,whatshouldbedonetoreachtherequirements.Whenacertainapplicationormarketis found, then the sensorshouldbeoptimized todetermine the limitsof thedevice.Thissearchprocess isan interactiveprocess.That is, technologicaloptionsgenerateideasonapplications,andviceversa.Thediscussionfocusedprimarilyontechnologyandperformance,whereapplicationandmarketspecificconsiderationsandespeciallysocietal embedding considerations were backgrounded. Thinking on themulti‐pathmap started from the technology side rather than the application side, which isunderstandableconsideringthatfewparticipantswerepresentwhocoulddiscussanapplicationfield.Thiswasintentionalinthedesignofthemeetingastheapplicationareas for this device were still very open. From this discussion we analysed theperformancecharacteristicsofthedeviceandinwhichapplicationsandmarketsthiswouldbeseenasuseful.Asecondroundcouldbeorganisedwhereactorsfromthesemarketsarepresentandcanthusstimulatemorethinkingfromtheapplicationside.


The interactions during the workshop helped to put some prioritization in thepathwaysidentifiedinadvanceandthosethatwereaddedlaterduringtheworkshop.Theidentificationofapplicationshadthecharacterofopening‐up(moreapplications)and closing‐down articulation processes (reducing amount of applications). Thisoccurred mainly from anticipating future performance of the device and assessingwhereitcouldmakeadifference.Whilefewconcreteapplicationswereidentified,thecharacteristics of such applications became increasingly clear (size‐selectivity andrapiddetectionmoreimportantthansensitivity).

PersonalreflectionontechnologyassessmentThe TA workshop was a useful tool for me to get a broader look on my researchproject. The discussionswith different people from different fields resulted in newideas and a clearer view on the future perspectives of the YI using multiplewavelengthstomeasuresize‐selectively.Bytakingthehelicopterviewitwaseasiertodeterminewhataremostimportantresultsandwhatistheaddedvalueofthissensor.Furthermore, I got a clearer view on possible applications and which researchquestions are important to answer for these applications. Tomake a TAworkshopmore useful in general, I think it is better to do such an exercise closer to thebeginning(beginningsecondyear)oftheprojectandnotinthefinalyearasIdid,suchthat the output of the TA exercise can help to guide the researcher in the rightdirectionandmakeshim/hermoreawareofthepossibledirectionsheorshecouldorshouldtake.Finally,itshouldbenotedthattheTAexercisetakesalotoftimetodoitproperly. If PhD’s are asked to do this, this time should be given to them and alsosufficienthelpfromTApeopleisrequired.WhentheTAexerciseisdoneproperly,thistime can be regained, because it is easier to see where the opportunities lie andthereforebetterdecisionscanbemade.

Acknowledgements

IwouldliketothankDr.HaicoteKulveforhishelpinpreparing,organisingandpost‐processingoftheworkshopandforhisassistanceonTA.Furthermore,thankstoDr.DaanSchuurbiersforhisassistanceonRATAandbringingmeincontactwithHaico.AlsoIwouldliketothankDr.VerenaStimbergforchairingandpost‐processingoftheworkshop.Iwouldliketothankalltheparticipantsoftheworkshopforsharingtheirknowledge and for their input during the active discussion in the workshop.Additionally,IwouldliketothankDaanSprünkenforhiscriticallookandcommentson theoutcomesof theworkshop.Thiswork is supportedbyNanoNextNL, amicroand nanotechnology consortium of the Government of the Netherlands and 130partners.

166| Chapter8

Appendix8.AWorkshoppreparations

Fig.8.A.1:Firstdraftof themulti‐pathmap illustratingpossibleapplicationsof size‐selectivedetection.


Table8.A.1:Groupcomposition3Participant Function

RonGill ExpertbiosensorsChristianBlum Measurementexpert,temporarysupervisorHarmenRoyKolkman BusinessdevelopmentUTwente,DirectorHightechfactoryDaanSprünken Physicist,OstendumR&Db.v.,PAImagingVinodSubramaniam Expertbiophysics,directorAMOLF,promotorHarmenVerenaStimberg Chair,TAresearcherHaicoteKulve Organizer,TAresearcherHarmenMulder Organizer,PhDresearchprojectYounginterferometer

Table8.A.2:WorkshopprogramTime What Goals Comments

10:00‐10:10

Welcome,introductionround

‐Everyonegetstoknowbackgroundofparticipants

‐WelcomebyHarmen‐Announcegoaloftheworkshop‐MappingexercisebasedoncooperationwithVerenaandHaicoofSTEPS(Science,Technology&PolicyStudies)‐Emphasizeinformality:asmuchdiscussionaspossible,brainstormcharacter,askingunderlyingideasandconsiderations‐Verenawillchairtheworkshopandwillwatchoverthetime‐Introductionround

10:10‐10:30

PresentationHarmenabout:‐research(introductiontoresearch)‐goalsofworkshop

‐Explainresearchtoparticipantssuchthattheyknowenoughaboutittoparticipateinthediscussion‐Informparticipantsonprogramandgoalsofworkshop‐Introducemulti‐pathmap(seeFig.8.A.1)

‐Participantsareallowedtoaskquestionsforclarification

10:30‐10:55

Identifyingapplications(makeuseofmulti‐pathmap,post‐itnotescanbeaddedbyparticipant/organizer)

‐IdentifypossibleapplicationsifmultiplewavelengthsareusedtoimprovespecificityofYIsensor‐IdentifypossibleapplicationsofYIsensorwithmultiplewavelengthsnexttousingitforimprovementofspecificityofhetsensor

‐5minutes:participantswritedownonpost‐itnote:functionality,product,market.Oneperpost‐itnote.Ortheymakepost‐itnottochangethecurrentmulti‐pathmap.‐Round:participantsexplainpost‐itnote.Otherparticipantsareaskedtorespondonthis,asktoassumptions,underlyingthoughtsabouttheassessment.Possiblequestionstoparticipants:‐Arefunctionalities/products/marketsrelevant?Whyorwhynot?‐Arefunctionalities/products/marketsdesirable?Whyorwhynot?‐Issensingofdiffusionpossibleandisthisinteresting?

3 Participantsagreedwithmentioningtheirnames

168| Chapter8

Time What Goals Comments

10:55‐11:35

Assessinginnovationpathways(makeuseofmulti‐pathmap,post‐itnotescanbeaddedbyparticipant/organizer)

‐Identifyaccompanyingpathwaysofpossibleapplications(fromtechnologytomarket(possiblysocietalembedding))‐Assesspathways:identifytechnological/commercialandbroaderchallenges/opportunitiesinaligningtechnologies–functionalities–products–markets–societalembedding.‐Iterations:identificationofnewapplications,newpathwaysandchallenges

‐5minutes:participantswritedownonpost‐itnote:pointsofinterestandsolutionsoffunctionality/product/market‐Round:participantsexplainpost‐itnote.Otherparticipantsareaskedtorespondonthis,asktoassumptions,underlyingthoughtsabouttheassessment.Possiblequestionstoparticipants:‐Arefunctionalities/products/marketsrealisable?Whyorwhynot?‐Howcanitberealised?

11:35‐11:50

Implicationsofassessment(makeuseofmulti‐pathmap:arrows,exclamationmarks)

‐Identificationofmostplausibleandpromisingpathways‐Identificationofnextstepsinresearch‐Identificationofpossiblecollaborations?

Possiblequestionstoparticipants:‐Whatarethemostpromisingroutes?‐Whathastobedonetorealisethis?‐Whatisthevisionoftheexperts?‐Whatisthevisionofthebusinessdeveloper?

11:50‐12:00

Conclusions

‐writedownmostimportantconclusions‐summarizeanswertoresearchquestions‐thankparticipants

‐Whatarethemostimportantfindings?‐Whataremoststrikinginsights?


Appendix8.BPicturesworkshop

170| Chapter8

References

1. A.Brandenburg,R.Krauter,C.Kunzel,M.Stefan,andH.Schulte,“Interferometricsensorfordetectionofsurface‐boundbioreactions,”Appl.Opt.39,6396‐6405(2000).

2. A.Ymeti,J.S.Kanger,R.Wijn,P.V.Lambeck,andJ.Greve,“Developmentofamultichannelintegratedinterferometerimmunosensor,”Sens.Actuators,B83,1‐7(2002).

3. M.C.Coen,R.Lehmann,P.Groning,M.Bielmann,C.Galli,andL.Schlapbach,“AdsorptionandbioactivityofproteinAonsiliconsurfacesstudiedbyAFMandXPS,”J.ColloidInterfaceSci.233,180‐189(2001).

4. S.Ohnishi,M.Murata,andM.Hato,“CorrelationbetweenSurfaceMorphologyandSurfaceForcesofProteinAAdsorbedonMica,”Biophys.J.74,455‐465(1998).

5. A.Rip,“TechnologyAssessment,”inInternationalEncyclopediaoftheSocial&BehavioralSciences(SecondEdition),J.D.Wright,ed.(Elsevier,Oxford,2015),pp.125‐128.

6. J.SchotandA.Rip,“Thepastandfutureofconstructivetechnologyassessment,”Technol.Forecast.Soc.Change54,251‐268(1997).

7. A.RipandH.teKulve,“ConstructiveTechnologyAssessmentandSocio‐TechnicalScenarios,”inPresentingFutures,E.Fisher,C.Selin,andJ.Wetmore,eds.(SpringerNetherlands,2008),pp.49‐70.

8. D.K.R.RobinsonandT.Propp,“Multi‐pathmappingforalignmentstrategiesinemergingscienceandtechnologies,”Technol.Forecast.Soc.Change75,517‐538(2008).

9. R.Phaal,C.J.P.Farrukh,andD.R.Probert,“Technologyroadmapping—Aplanningframeworkforevolutionandrevolution,”Technol.Forecast.Soc.Change71,5‐26(2004).

10. “Optiqua”,retrieved2015,http://www.optiqua.com/mach_zehnder.html.11. “ChemSpider”,retrieved2015,http://www.chemspider.com/Chemical‐

Structure.4575379.html?rid=3160d5fc‐8f52‐40cc‐882e‐7aa6ea02d103,http://www.chemspider.com/Chemical‐Structure.10194104.html.

12. I.Bjork,B.A.Petersson,andJ.Sjoquist,“SomephysiochemicalpropertiesofproteinAfromStaphylococcusaureus,”Eur.J.Biochem.29,579‐584(1972).

Listofabbreviations

Abbreviation MeaningBiMW BimodalwaveguideCCD Charge‐coupleddeviceCT CrosstalkCTA ConstructivetechnologyassessmentDLS DynamiclightscatteringDPI DualpolarizationinterferometerDPSS Diode‐pumpedsolidstateELISA Enzyme‐linkedimmunosorbentassayFFT FastFouriertransformIO IntegratedopticalIP IntellectualpropertyLED Light‐emittingdiodeLOD LimitofdetectionMZI Mach‐ZehnderinterferometerNE EffectiverefractiveindexerrorPBS Phosphate‐bufferedsalinePCR PolymerasechainreactionPE Phaseerrorpk‐pk Peak‐to‐peakPTC PhotontransfercurveQ‐factor QualityfactorRI RefractiveindexRIU RefractiveindexunitsRMS RootmeansquareSEM ScanningelectronmicroscopeSNR Signal‐to‐noiseratioSP SurfaceplasmonSPR SurfaceplasmonresonanceTA TechnologyassessmentTE TransverseelectricTIR TotalinternalreflectionTM TransversemagneticWGM WhisperinggallerymodesWIOS WavelengthinterrogatedopticalsensorYI Younginterferometer

SummaryThis thesis presents a Young interferometer (YI) biosensor that can perform size‐selective measurements using multiple wavelength excitation. This size‐selectivedetection can be used to improve the specificity of evanescent field‐based opticalbiosensorsandisbasedonthevarioussensitivitiesoftheevanescentfieldsofmultiplewavelengths. The approach of using multiple wavelengths is, in addition to YIbiosensors,alsoapplicabletoothertypesofevanescentfield‐basedopticalsensors.InChapter 1 we first introduce the general concept of biosensors, followed by adiscussion of important criteria for biosensors. The most widely‐used evanescentfield‐basedoptical sensorsarereviewed. The limitingspecificityof thesesensors isaddressed,togetherwithtechniquestoimprovethespecificity.Finallyweproposethenewapproachofsize‐selectivedetection,basedontheuseofmultiplewavelengthstoimprovethespecificityofevanescentfield‐basedopticalbiosensors.

InChapter2wepresentasize‐selectivedetectionmethod for integratedopticalinterferometric biosensors that could strongly enhance their performance. WedemonstratethatbylaunchingmultiplewavelengthsintoaYIwaveguidesensoritisfeasible to derive refractive index changes ( n ) from different layers above thewaveguidesurface,enablingonetodistinguishbetweenboundparticles(e.g.proteins,viruses, bacteria) based on their differences in size and simultaneously eliminatinginterferencefroma n inthebulk(regionofafewhundrednanometresabovesensorsurface).Numericalcalculationsareusedtooptimizesensordesignandthedetectionmethod. Adding size‐selectivity to the sensor reduces the sensitivity of the sensor.However, the theoretical sensitivity remains still comparable to other existingbiosensors when discriminating between n ’s in three different layers above thewaveguide based on simultaneous detection of effective refractive index changes ( effN ) at three different wavelengths. Assuming a particle of 80 nm in size as the

specific analyte to detect, the theoretically determined minimum detectable masscoverage is 4×102 fg/mm2 (assuming a phase noise of 10‐4 fringes). This isapproximately one order of magnitude higher than the minimum detectable masscoveragewiththeYIusingasinglewavelength.However,withsize‐selectivedetectionit isnowpossibletodiscriminatethe80nmsizedanalytebindingfromnon‐specificboundparticlesof10nmsizeandsimultaneouslyoccurringbulkchanges.

Chapter 3 presents the design, realization, and characterization of a YI sensorsetupcapable tomeasuresimultaneously effN ’satmultiplewavelengthstoachieve

size‐selectivity. First, the requirementsandanoverviewof the setuparepresented.Next,wedescribetherealizationandcharacterizationofthesixmainmodulesofthesetup: light sources, incoupling, sensing platform, imaging, detection and dataprocessing. Finally, we characterise the phase noise and phase drift (unexpected

174 Summary

phase changesona longer time scale)of the setup. Itwas found that themeasuredphasenoiseonshort timescales isdeterminedbyphotonshotnoise.ACCDcamerawith a high dynamic range was implemented in the setup such that themeasuredphase noise is smaller than 10‐4 fringes@ 1 Hz. Phase drift is smaller than 5x10‐3fringesper1000sforthesetupwithend‐firecoupling.Theincouplingwasmademoretime efficient by using a fiber and butt‐end coupling. However, the phase driftincreased to values smaller than 1.5x10‐2 fringes per 1000 s, partly caused bypositionaldriftofthefiber.Driftindistancebetweenfiberandchipresultedinphaseoscillationswhichcouldbesolvedusingamatchingindexgel. Itwasalsofoundthatairflowcanstronglyinfluencethephasesignal,sothesetupwascovered.Moreover,changesintheambienttemperaturehaveaneffectonthephasestabilityasitcanleadtothermalexpansionofcomponentsofthesetupandthechipitself.Thiscanleadtochangesinthepositionofthechipwithrespecttothecamerawhichresultsinphasechanges. Next to drift also artefacts are observed in the phase signal. The origin ofthese artefacts is investigated in this chapter. Possible reasons of the artefacts areaberrationsofgratingandlensesintheimagingpartofthesetupandboundaryeffectsoftheshiftinginterferencepattern.Theartefactsshowupasoscillationsinthephasechange with an amplitude in the order of 10‐2 fringes. Induced signals should besignificantly higher than drift and artefacts to minimize the influence of drift andartefactsonthedetermined n ’s.

In Chapter 4 we present a description and a detailed study of different signalanalysis approaches that are required to obtain the signal from the differentsubstances from the measured phase changes at the different wavelengths. It wasfound that a theoretical approachaspresented inChapter2 is exactbut inpracticedifficult to implement because of the many parameters that have to be tuned.Therefore,wealsodevelopedamuchmorepracticalratio‐basedapproachbasedontheratiosof effN ’smeasuredatdifferentwavelengths.Theseratiosweredetermined

independently for 85 nm carboxylated polystyrene beads (representing specificbinding of e.g. viruseswhichhave approximately this size), proteinA (representingnon‐specific binding) and D‐glucose (representing bulk changes). These ratios areusedtodiscriminatebetweenthe n ’sinducedbythesesubstanceswhichareappliedintheproof‐of‐principleexperiments forsize‐selectivedetection.Ontheotherhand,using this ratio‐basedapproach it is not possible to determine an absolute value ofn .However,this isnotstrictlyrequired,becausecalibrationmeasurements,which

are usually also carried out by a singlewavelength YI, can be used to calibrate themeasuredvaluesof n withananalyteconcentration.Thetheoreticalandratio‐basedapproach were combined to determine the absolute value of n and thecorresponding surfacemass coverage.Moreover, the approacheswere combined todeterminethethicknessesof85nmbeadstobe≈72.6nmandproteinAtobe≈2.1nm which are in agreement with literature. Furthermore, the influence of the

Summary 175

parameters that are required for the analysis approaches (e.g. waveguide corethickness, waveguide refractive indices, ratios of effN ’s) on the determined n ’s

causedindifferentlayersorbydifferentsubstanceswastested.Finally,simulationsofa real experimental setting confirm the theoretical analysis given in Chapter 2 thatnoise,driftandartefactsshowupenhancedinthedetermined n ’s.

Inchapter5wepresenttheexperimentalapplicationofthesize‐selectiveanalyteapproaches presented in Chapter 4. We use the different analysis approaches todiscriminatebetweenbindingof85nmbeadsfrombindingofproteinA,andbindingof85nmbeadsorproteinAfromD‐glucosebulkchangesbymeasuring effN attwo

wavelengths.Threewavelengthswereusedtodiscriminatebetweenthreesubstancessimultaneously, however this led to noisy results which leaves the applicabilityuselessatthismoment.Forasuccessfulapplicationoftheseanalysisapproachesusedtodeterminethreeindependent n ’s,theaforementionedartefactsanddriftin effN

shouldbereducedsignificantly.Alternatively, itwasshownthatothertechniquestoimprove specificity, such as the use of a reference channel to compensate for bulkchanges,canbeusednexttosize‐selectivedetection.Theworkingoftheratio‐basedapproachtodiscriminatebetweentwosubstanceswasverifiedbyablindexperimentwith samples containing different concentrations of protein A and 85 nm beads.Furthermore,a triplicateofmeasurementsserieswithdifferentbeadconcentrationsand a constant protein A concentration showed that we could discriminate n ’scausedbybindingofproteinAand85nmbeads,evenwhenthe n oftheproteinwasapproximately20timeshigher.Thiscanforexamplebeusedtodiscriminatespecificanalyte binding of larger particles from non‐specific binding of smaller particles toimprovetheperformanceoftheYIsensorandIOinterferometricsensorsingeneral.

Until now, size‐selective detection with the YI sensor was done using multiplewavelengthsbasedontheirdifferentmodeprofilesandtheirdifferentsensitivitiesindifferentlayersintheevanescentfield.Chapter6showsthatinasimilarway,multiplepolarizations can also be used to discriminate between n ’s of multiple layers ormultipledifferent‐sizedsubstances.Calculationsfromchapter2showedthatthemoredifferentthesensitivitycoefficientsofeachmode,thelesssingularthematrixtosolven ’s in multiple layers or induced by multiple substances, meaning that all noise

sources in effN show up less enhanced. We show that it is possible to measure

multiplepolarizationsandmultiplewavelengthssimultaneously,resultinginamatrixwhich is lesssingularforsomecasesandthereforerepresentingan improvementofthe approach in these cases. However, it should be noticed that measuringsimultaneouslywithmultiplewavelengthsandpolarizationsalsoledtoanincreaseoftheaforementionedartefactsinthemeasurements.Forasuccessfulapplicationoftheuse of multiple polarizations and wavelengths for size‐selective detection, theartefactsinthemeasurementsshouldbereducedsignificantly.

176 Summary

Fromthedetermined n ’swesawthatthe n duetobindingofproteinAtothesurfaceoccurredearlier than the n due toproteinA in thebulk.However, itwasexpectedthatproteinAshouldarriveinthebulkbeforeincanattachtothesurface.InChapter7thisunexpectedtimedelayisdiscussed.Finiteelementmethodsimulationson the diffusion and binding ofmolecules show that a time delay between surfacebindingofproteinAandproteinAbulkchangescanarisewhenproteinAhasahighaffinity for the surface.Thebulk region isdepleteddue to the fact thatproteinA isstrongly attracted towards the surface. The timedelay increaseswith an increasingmaximum surface capacity max and increasing adsorption rate ka. For a ka of 105

m3/mol/s and a max between 4.4x10‐3 and 1.7x104 fg/mm2 (corresponding to a

protein A radius of 1 and 2 nm respectively which agrees well to the measuredthicknessof2.1nm(seeChapter4)),a timedelayofapproximately20s to100s isfound. This is comparable to the observed time delay. To verify the results of thesimulations,proteinAandD‐glucosewereaddedsimultaneouslytothesensor.DuetothefactthatD‐glucosehasamuchhigherdiffusionthanproteinAitwasexpectedthatn due to D‐glucose bulk changes would appear earlier now. However, again the

signal of n due to binding of protein A occurred earlier than the n due to D‐glucose.Experimentshintthatthistimedelayisrelatedtothetransporttowardsthesensingwindowas increasing the transporting tube lengthresulted inan increasingtimedelay.Experimentsshowedthatthe timedelayalsooccurred forbovineserumalbumin,suggestingthatthetimedelayisindependentoftheproteinused.However,theexactoriginofthetimedelaybetweenD‐glucosebulkchangesandsurfacebindingofproteinAisnotfound.

Chapter 8 explores what are the most relevant applications for the Younginterferometer (or a similar) sensor in combination with size‐selective detectionbasedonmultiplewavelengthsand/orpolarisations.Inthischapterwealsoaddresswhatshouldbenextstepsinfutureresearchbasedontheconclusionsdrawnfromtheotherchapters.Forthatpurpose,asmalltechnologyassessmentstudywasperformedand a mini‐workshop was organised to identify possible innovation pathways andrelevant factors to consider from a technological, product and commercializationpoint of view. From the assessment it was determined that the added value of thesensoristomeasuresize‐selectively(discriminatebetweensubstancesbasedontheirsize)and that thegain in size‐selectivity shouldbeworth the loss in sensitivityandstability and the increasing costs. Therefore, pharmacy and the medical sector areprobably less suitablemarkets for this sensor, as in theseapplicationareas, specificbiochemistry and sensitivity are more important than size‐selectivity. A researchplatform foruniversitiesor companieswas found tobe themostpromisingmarket,but it must be noticed that markets should be analysed inmore depth for amorecomplete picture. Possible applications in this market are measurement tools tomeasure the kinetics of layer growth, a signal enhancer, a quick screener or an

Summary 177

alternative for a Scanning Electron Microscope, an alternative for Dynamic LightScattering (tomeasure thicknessof layersorparticles)anda research tool tostudyfundamental research on binding or other surface effects or flow dynamics. Thisresultedinresearchquestions“Whataretheminimumandmaximumsizewhichcanbedetectedanddetermined?”and“Whatdifferences insizescouldbediscriminated fromeachother?”.Furtherresearchisrequiredtoanswerthesequestions.Whenacertainapplicationormarketisidentified,thenthesensorshouldbeoptimizedtodeterminethelimitsofthedeviceforthespecificapplicationinmind.

SamenvattingIn dit proefschrift wordt een Young interferometer (YI) gepresenteerd die selectiefkan meten op basis van grootte, gebruikmakende van detectie met meerderegolflengtes. De selectiviteit op basis van grootte kan worden gebruikt om despecificiteit van optische biosensoren (die gebruik maken van detectie met hetevanescente veld) te verbeteren en is gebaseerd op de verschillende gevoelighedenvan de evanescente velden van de verschillende golflengtes.De benadering van hetgebruikvanmeerderegolflengtesisnaastdeYIooktoepasbaaropanderetypesvanevanescente‐veld‐gebaseerde optische sensoren. In hoofdstuk 1 introduceren weeerst het algemene concept van biosensoren, gevolgd door een discussie van debelangrijkste criteria van biosensoren. De meest gebruikte evanescente‐veld‐gebaseerde sensoren worden besproken. Daarnaast wordt de gelimiteerdespecificiteit van deze sensoren aan de orde gesteld, samen met technieken diegebruikt kunworden omde specificiteit te verbeteren. Ten slottewordt de nieuwebenadering van selectiviteit op basis van grootte (grootte‐selectieve detectie)gebaseerdopdetectiemetmeerderegolflengtesvoorgedragenomdespecificiteitvanevanescente‐veld‐gebaseerdeoptischesensorenteverbeteren.

In hoofdstuk 2 presenteren we een grootte‐selectieve detectiemethode voorgeïntegreerde optische interferometrische biosensoren die hun presentaties sterkkunnen verbeteren. We demonstreren dat het mogelijk is ombrekingsindexverschillen ( n ) te bepalen in verschillende lagen boven hetoppervlaktevandegolfgeleiderdoormiddelvandetectiemetmeerderegolflengtes.Hiermee maken we het mogelijk om gebonden deeltjes (bijvoorbeeld eiwitten,virussen, bacteriën) te onderscheiden van elkaar op basis van grootte en dittegelijkertijd te onderscheiden van een homogene n in het complete evanescenteveld (bulk). Numerieke berekeningen zijn gebruikt om het sensorontwerp en dedetectiemethode te optimaliseren. Het toevoegen van grootte‐selectiviteit leidt totreductievandegevoeligheidvandesensor.Detheoretischegevoeligheidblijftechternog steeds vergelijkbaarmet de gevoeligheid van andere vergelijkbare biosensorenwanneerweonderscheidmakentussen n ’svandrieverschillendelagen,gebaseerdop simultane detectie van drie verschillende golflengtes. Als we aannemen dat dedeeltjesdiewedetecteren80nmgrootzijn,danisdetheoretischbepaaldeminimaaldetecteerbare oppervlaktemassadekking 4×102 fg/mm2 (een faseruis aannemendevan 10‐4 fringes). Dit is ongeveer één ordegrootte hoger dan de minimaaldetecteerbare oppervlaktemassadekking van de YI op basis van detectie met eenenkele golflengte. Echter met grootte‐selectieve detectie is het nu mogelijk om debindingvande80nmgroteanalytenteonderscheidenvannon‐specifiekebindingvandeeltjesvan10nmgrootentegelijkertijdplaatsvindendebulkveranderingen.

180 Samenvatting

Hoofdstuk3presenteerthetontwerp,derealisatieendekarakteriseringvandeYIsensor opstelling die het mogelijk maakt om simultaan effectievebrekingsindexverschillen ( effN ) te detecteren bij verschillende golflengtes om

grootte‐selectief te kunnen meten. Eerst worden de eisen en een overzicht van deopstellinggepresenteerd.Daarnabeschrijvenwederealisatieenkarakteriseringvande zes belangrijkste onderdelen van de opstelling: de lichtbronnen, de inkoppeling,hetsensorplatform,hetafbeelden,dedetectieendegegevensverwerking.Tenslottekarakteriseren we de faseruis en fasedrift (onverwachte faseveranderingen op eengrotere tijdschaal) vandeopstelling.De gemeten faseruis op korte tijdschaalwordtbepaald door foton‐hagelruis. Een CCD‐camera met een hoog dynamisch bereik isgeïmplementeerd in de opstellingen zodat de gemeten faseruis kleiner is dan 10‐4fringes. De fasedrift is gebruikelijk lager dan 5x10‐3 fringes over 1000 s voor deopstellingmetinkoppelingvanhetlichtindegolfgeleiderviaeenlens.Deinkoppelingis tijdefficiënter gemaakt door het licht in te koppelen via een fiber. Echter debijkomende fasedrift nam toe tot waardes lager dan 1.5x10‐2 fringes over 1000 s,gedeeltelijkbepaalddoorhetlangzaamwegdriftenvandefiber.Driftinafstandtussende fiberendechipresulteerde in faseoscillaties.Ditprobleemkonopgelostwordendoor gebruik te maken van een matching‐index‐gel. Verder is gemeten datluchtstromingenhetfasesignaalsterkkunnenbeïnvloeden,waardoorwedeopstellingvolledighebbenbedekt.Tevenshebbenomgevingstemperatuurveranderingeninvloedopdefasedrift,omdathetkanleidentotthermischeuitzettingvancomponentenindeopstellingendechipzelf.Ditkanleidentotveranderingindepositievandechiptenopzichtevandecamera,watresulteert in faseveranderingen.Naastdedriftzijnookartefacten gemeten in het fasesignaal. De oorsprong van deze artefacten zijn in dithoofdstukonderzocht.Mogelijkeoorzakenzijndeafwijkingenvandetralieoflenzentussen de chip en de camera of randeffecten van een verschuivendinterferentiepatroon. De artefacten zijn te zien als oscillaties in het fasesignaalmeteen amplitude in de ordegrootte van 10‐2 fringes. Geïnduceerde signalen moetensignificant hoger zijn dan de drift en de artefacten om de invloed van de drift enartefactenopdebepaalde n ’steminimaliseren.

Inhoofdstuk4presenterenwedebeschrijvingeneengedetailleerde studievanverschillende analysemethodes die nodig zijn voor het verkrijgen van de n ’s vanmeerdere substanties/laagjes op basis van de gemeten faseveranderingen metmeerdereverschillendegolflengtes.Hetisbevondendatdemethodevanhoofdstuk2exact is,maar indepraktijkmoeilijk te implementerenvanwegehet feitdaterveleparameters afgesteld moeten worden. Daarom hebben we een veel praktischeremethodeontwikkeldgebaseerdopderatio’svande effN gemetenbijverschillende

golflengtes. Deze ratio’s zijn afhankelijk gemeten voor 85 nm gecarboxyleerdepolystyrenebolletjes(representerenspecifiekebindingvanbijvoorbeeldvirussendieongeveer deze grootte hebben), het eiwit proteïne A (representeert non‐specifieke

Samenvatting 181

binding)enD‐glucose(representeertbulkveranderingen)zodatdezeratio’sgebruiktkunnenwordenomonderscheid temaken tussendeze substantiesdiegebruikt zijnvoordeexperimentenomaantetonendatgrootte‐selectievedetectiewerkt.Aandeandere kant is het niet mogelijk om, op basis van de ratio‐gebaseerde methode,absolute waarden van n te meten. Dit is echter ook niet noodzakelijk, omdatkalibratiemetingen, die normaal ook gebruikt worden bij detectie met een enkelegolflengte, gebruikt kunnenwordenomde gemetenwaarden van n te kalibrerenmeteenanalytconcentratie.Detheoretischemethodeenderatio‐gebaseerdemethodezijn ook gecombineerd om de absolute waarden van n en de corresponderendeoppervlaktemassadekking te bepalen. Verder zijn dezemethodes gecombineerd omdegroottevande85nmbolletjestebepalenals≈72.6nmenproteïneAals≈2.1nm,wat overeenkomtmet de literatuur. Verder is de invloed van inputparameters, dienodig zijn voor de analysebenaderingen (bijvoorbeeld golfgeleiderkerndikte engolfgeleiderbrekingsindices, ratio’s van effN ), op de bepaalde n ’s van de

verschillende laagjes of substanties getest. Ten slotte bevestigen simulaties van eenechteexperimentelesettingdatruis,driftenartefactenversterkttevoorschijnkomenindebepaalde n ’s.

In hoofdstuk 5 presenteren we de experimentele toepassing van debovengenoemdeanalysemethodesvoorgrootte‐selectievedetectie.Wegebruikendeverschillende analysemethodes om te onderscheiden tussen de binding van 85 nmbolletjes en de binding van proteïne A en om D‐glucose bulkveranderingen teonderscheidenvandebindingvan85nmbolletjesofdebindingvanproteïneA,door

effN te meten bij twee golflengtes. Drie golflengtes worden gebruikt om te

onderscheiden tussen alle drie substanties tegelijkertijd. Dit leidde echter totresultaten waarin de ruis versterkt zichtbaar werd, waardoor grootte‐selectievedetectie,omteonderscheidentussen n ’svandriesubstanties/laagjesopbasisvandrie golflengtes, met de huidige opstelling op dit moment onbruikbaar is. Voorsuccesvolleapplicatievandeanalysemethodesomdrie n ’steonderscheiden,zulleneerdergenoemdeartefactenin effN endriftsignificantgereduceerdmoetenworden.

Als alternatief is het laten zien dat het ook mogelijk is om andere technieken omspecificiteitteverbeteren,zoalshetgebruikvaneenreferentiekanaalombulkeffectentecompenseren,tegebruikennaastdegrootte‐selectievedetectie.Dewerkingvanderatio‐gebaseerdemethodeisgeverifieerdmeteenblindexperimentmetverschillendemonstersbestaandeuitverschillendeconcentratiesvanproteïneAen85nmbolletjes.Daarnaast liet een drievoud van meetseries met verschillende 85 nm bolletjesconcentraties en een constante proteïne A concentratie zien dat we kondenonderscheidentussen n ’sveroorzaaktdoordebindingvanbolletjesendebindingvan eiwitten, zelfs als de n van het eiwit 20 keer zo hoogwas als de n van debolletjes.Ditkanbijvoorbeeldwordengebruiktomteonderscheidentussenspecifiekeanalytbindingvangroteredeeltjesennon‐specifiekebindingvankleineredeeltjesom

182 Samenvatting

depresentatiesvandeYIsensorengeïntegreerdeoptischesensoreninhetalgemeenverbeteren.

Totnu toewerddegrootte‐selectievedetectiemetdeYI sensoruitgevoerdmetmeerderegolflengtesgebaseerdophunverschillendegevoelighedeninverschillendelagen inhetevanescenteveld.Hoofdstuk6 laatzietdatopeenvergelijkbaremanierookverschillendepolarisatiesgebruiktkunnenwordenomteonderscheiden tussenverschillende n ’s. Berekeningen van hoofdstuk 2 lieten zien dat hoe meer degevoeligheden van de verschillende golflengtes van elkaar verschillen, hoe mindersingulierdematrix(dienodigisom n ’stebepalen)is.Ditbetekenddatderuisinde

effN ’s minderversterkt tevoorschijnkomt indebepaalde n ’s.We latenziendat

het mogelijk is om meerdere polarisaties en golflengtes simultaan te meten,resulterende ineenmindersingulierematrixvoorsommigegevallenendaaromeenverbeteringvandebenaderingindezegevallen.Hetmoetechterwordenopgemerktdat het simultaan meten van meerdere golflengtes en polarisaties leidde tot eentoenamevanartefactenindemetingen.Vooreensuccesvolletoepassingvangrootte‐selectieve detectie op basis van verschillende polarisaties en golflengtes zullen deartefactenindemetingensignificantgereduceerdmoetenworden.

Uitdebepaalde n ’s zagenwedatde n doorbindingvanproteïneAaanhetoppervlakte eerder plaats vonddande n door de proteïneA in de bulk.HetwasechterverwachtdatdeproteïneAeerstindebulkzoumoetenarriverenvoordathetkan binden aan het oppervlakte. In hoofdstuk 7 wordt deze onverwachtetijdsvertragingbediscussieerd.Eindige‐elementenmethode‐simulatiesnaardediffusieen binding van moleculen laten zien dat de tijdsvertraging kan ontstaan wanneerproteïneAeenhogeaffiniteitheeftmethetsensoroppervlakte.DoordatdeproteïneAsterk wordt aangetrokken door het oppervlakte zal er depletie van de bulkregioplaatsvinden. De tijdsvertraging neemt toe met een toenemende maximaleoppervlaktecapaciteit max eneentoenemendeadsorptieconstanteka.Vooreenkavan

105m3/mol/s en een max tussen 4.4x10‐3 en 1.7x104 fg/mm2 (komenovereenmet

respectievelijkeenproteïneAradiusvan1en2nm,watongeveerovereenkomtmetdebepaalddiktevan2.1nminhoofdstuk4)isereentijdsvertragingvan20stot100sgevonden.Dezetijdsvertragingenzijnvergelijkbaarmetdegemetentijdsvertragingen.Om de resultaten van de simulaties te verifiëren, zijn D‐glucose en proteïne Ategelijkertijd aan de sensor toegevoegd. DoordatD‐glucose een veel hogere diffusieheeft dan proteïne A, werd verwacht dat de n door D‐glucose eerder zouplaatsvinden. Echter opnieuw vond de n door de binding van proteïne A eerderplaatsdande n vanD‐glucose.Experimentenhintenernaardatdevertragingwordtveroorzaak tijdens het transport naar het sensorgebied toe, omdat de vertragingtoeneemt naarmate de slanglengte van het transport werd vergroot. Ook latenexperimenten zien dat de tijdsvertraging optreedt bij het eiwit bovine serumalbumine, suggererende dat de tijdsvertraging onafhankelijk is van het gebruikte

Samenvatting 183

eiwit.EenpreciezeoorzaakvandetijdsvertragingtussenbulkveranderingendoorD‐glucoseendebindingvanproteïneAisechternietgevonden.

Hoofdstuk 8 zoektwat demeest relevante toepassingen zijn voor de YI (of eenvergelijkbare) sensor in combinatie met grootte‐selectieve detectie gebaseerd opdetectiemetmeerdere golflengtes en/of polarisaties. Verder onderzoekenwe in dithoofdstukwatvervolgstappenzoudenmoetenzijninvervolgonderzoek,gebaseerdopde conclusies getrokken uit de vorige hoofdstukken. Hiervoor is een kleintechnologischaspectenonderzoekuitgevoerdeniseenmini‐workshopgeorganiseerd,ommogelijke innovatiepaden en relevante factoren op het gebied van technologie,productencommercialisatie te identificeren.Uitditonderzoek isvastgestelddatdetoegevoegdewaarde van de sensor is dat het grootte‐selectief kanmeten (dus kanonderscheidentussensubstantiesopbasisvanhungrootte)endatdewinstdiehierinte behalen valtmoet opwegen tegen het verlies in gevoeligheid en stabiliteit en detoename inkosten.Daaromzullende farmacieendemedischesectorwaarschijnlijkminder geschikte markten zijn, omdat in deze toepassingsgebieden specifiekebiochemie en gevoeligheid belangrijker zijn dan grootte‐selectiviteit. Eenonderzoeksplatform voor universiteiten of bedrijven bleek demeest veelbelovendemarkt, maar het moet in acht worden genomen dat markten dieper geanalyseerdmoetenwordenvooreencompleterbeeld.Mogelijketoepassingenindezemarktzijnmeetgereedschappenomkinetiekvanlaaggroottestebepalen,eensignaalversterker,een snelle scanner of een alternatief voor een rasterelectronmicroscoop (SEM).TevenskanhetgebruiktwordenalseenalternatiefvoorDynamicLightScattering(omgroottevanlagenofdeeltjestemeten)ofalseenonderzoeksmiddelomfundamenteelonderzoektedoennaarbinding,andereoppervlakte‐effectenofstromingsdynamica.Ditresulteerdeindevolgendeonderzoeksvragen:“Watzijndeminimaleenmaximalegroottediegedetecteerd enbepaald kunnenworden?” en “Wat zijnde verschillen ingroottedievanelkaaronderscheidenkunnenworden?”.Meeronderzoekisvereistomdeze vragen te beantwoorden. Wanneer een bepaalde toepassing of markt isgeïdentificeerd,zaldesensormoetenwordengeoptimaliseerdomvasttestellenwatdelimietenvanhetinstrumentzijnvoordezespecifieketoepassing.

ListofpublicationsPapersH.K.P.Mulder, A. Ymeti, V. Subramaniam, J.S. Kanger, Size‐Selective Detection in IntegratedOpticalInterferometricBiosensors,OpticsExpress,20(19),20934‐20950,2012H.K.P.Mulder, C. Blum, V. Subramaniam, J.S. Kanger, Size‐selective analyte detectionwith aYounginterferometersensorusingmultiplewavelengths,submitted

OralpresentationsMulder, H.K.P., Subramaniam V., Kanger, J.S., Size‐selective detection in integrated opticalinterferometric biosensors, Annual Dutch meeting on Molecular and Cellular Biophysics,Veldhoven,TheNetherlands,2011,June16Mulder,H.K.P., Ymeti, A., Dudia, A., Subramaniam, V., Kanger, J.S., Size‐SelectiveDetection inIntegratedOptical InterferometricBiosensors,MicroNanoConference'12,Ede,TheNetherlands,2012,December10Mulder,H.K.P., Subramaniam, V., Kanger, J.S.,Size‐selectivedetection inan integratedopticalYounginterferometerbiosensor,MESA+dag,Enschede,TheNetherlands,2013,September16Mulder,H.K.P.,Subramaniam,V.,Kanger,J.S.,BackgroundReductioninaYoungInterferometerBiosensor,AdvancedPhotonicsCongress,Barcelona,Spain,2014,July31Mulder, H.K.P., Subramaniam, V. & Kanger, J.S., Background Reduction in a YoungInterferometerBiosensor,NanoCity2014,Utrecht,TheNetherlands,2014,October27

PostersMulder,H.K.P., Ymeti, A., Subramaniam,V.&Kanger, J.S.,Developmentofamulti‐wavelengthintegrated Young interferometer biosensor, Annual Dutch meeting on Molecular and CellularBiophysics,Veldhoven,TheNetherlands,2011,October03‐04Mulder,H.K.P., Ymeti, A., Dudia, A., Subramaniam, V. & Kanger, J.S.,Development ofaMulti‐WavelengthIntegratedYoungInterferometerBiosensor,Biosensors2012,Cancun,Mexico,2012,May15‐18Mulder,H.K.P.,Ymeti,A.,Dudia,A.,Subramaniam,V.&Kanger,J.S.,Size‐SelectiveDetectioninanIntegrated Young interferometer Biosensor, Label‐free technologies, Amsterdam, TheNetherlands,2012,November01‐03Mulder,H.K.P.,Subramaniam,V.&Kanger,J.S.,Size‐SelectiveDetectioninanIntegratedOpticalYoung Interferometer Biosensor, MicroNanoConference ’13, Ede, The Netherlands, 2013,NovemberMulder,H.K.P.,Subramaniam,V.&Kanger,J.S.,Size‐selectiveDetectioninanIntegratedOpticalYoungInterferometerBiosensor,NanoCity2014,Utrecht,TheNetherlands,2014,October27Mulder, H.K.P., Subramaniam, V. & Kanger, J.S., How can size‐selectivity be added to anintegratedYounginterferometerbiosensor?,NanoCity2015,Amersfoort,TheNetherlands,2014,October27

DankwoordHet iseindelijkzover,na jarenvanhard,maaraangenaamwerken,komternueeneindeaanmijnpromotie.Opditproefschriftstaatmijnnaam,maarhetwaszekerniettotstandgekomenzonderdehulpvanveleanderemensendieikgraaghiervoorzouwillenbedanken.

TeneerstewilikgraagmijndagelijksebegeleiderHansbedanken.Ikvondhetergfijnommetjesamentewerken.Alsikevenvastliepkonikaltijdbijjouwkantoornaarbinnenlopenenhielp jemeweer opweg. Verder heb ik jouw kritische blik en vooral ook jouw aanwas vannieuweideeënerggewaardeerd.Jammerdatjetijdelijkuitderunningwas,maarhetisgoedomteziendatjemomenteelweersteedsmeeraanhetwerkbent.Fijndatjenajeterugkeertochweer de tijd nam om ons artikel kritisch te bekijken. Daarnaastwil ikmijn promotor Vinodbedanken.Bedanktdatjemedekansgafomdezepromotietedoen.Ondanksdatweelkaarnietheelvaakzagengafjemetochaltijdhetvertrouwen.Daarnaastwaardeerikhetookzeerdatjemealtijdbentblijvenhelpenensteunenondanksdatjehetergdrukhadmetdenieuwebanendiejekreeg.

VerderwilookgraagSylviabedanken.Jebenteengeweldigekrachtindegroepenjekuntaltijdbij jouterechtmeteenverhaalenvragenovervanallesennogwat.Bedanktvoordezegeweldigesteun.ChristianenRon,thankyouforhelpingmeduringthetimethatHanswasnotpresent.Christian,Igreatlyappreciateitthatyoutookthetimetoreadandcommentsomeofmychapters.OokMireilleenSaskiabedanktvoordesteunenhulpdiejulliemeopdatmomenthebbengegeven.AlsothankstothepeopleofourMondaymorninggroupmeeting.Hans,Peter,Burcu,Lantian,Ron,Kristianandallothers,thanksforyourinput,questionsandthediscussionswe had. Tevens wil ik graag de technici Kees en Robert bedanken. Als ik tegen problemenaanliepinhetlabkonikaltijdbijjullieterechtenjulliehielpenmebijdekarakteriseringenhetfabricerenvanonderdelenvoordeopstelling.OokwilikMartijnbedankendiemeopweghielpin het lab enmet het bestellen van de juiste componenten voor de opstelling. Erwin, ook jijbedanktvoorjouwhulpinhetbeginvanmijnproject.IkkonbijjouookaltijdaankloppenalsikvragenhadoverLabview.Daarnaastwil ikAurelenAlmabedankenvoorhet introduceren inhetwerkenmet de Young interferometer sensor en de dingen die hierbij komen kijken.Ookdank aan Xiophotonics, in het speciaal Ronald, voor de hulp bij het realiseren van defiberinkoppeling in de opstelling. In het biochemische lab kon ik voor vragen en hulp altijdterecht bij Yvonne, Kirsten, Irene en Nathalie: hartelijk dank hiervoor. Verder wil ik mijnkamergenotenArshdeep,Christian,Maurice,Rick,DodoenFedericabedanken,metwieikaltijdgoed op kon schieten.We konden altijd gezellig praten,maar er heerste ook bijna altijd eengoedewerksfeerwaarin iedereen rustig aan hetwerk kon. Chris, itwas also nice to be yourparanymph, tosharehotel rooms inconferencesand tomanage thewebsite together.VerderwilikalleandereNBPcollega’senex‐collega’sbedankenvoordegoedesfeerindegroependegezelligepauzes,julliewarenergleukecollega’s.

OokwilikHaicoenVerenabedankenvoorhunhulpaanhetTechnologyAssessmentdeelvanmijnproefschriftenhetopzettenenuitvoerenvandemini‐workshop.Zonderjulliegoedehulp en ondersteuning was me dit niet gelukt. Tevens wil de deelnemers van de workshopbedanken voor hun bijdrage en in het speciaal ook Daan: bedankt voor jouw input die jegegevenhebtvoordithoofdstuk.

188 Dankwoord

Next tomyPhDIalsohadagreat timeatP‐NUT.Therefore, Iwould liketo thankall thepeoplewho Iworkedwith and especiallymy colleague boardmembers Silja, Giovane, Bijoy,Juan, Juan Carlos, Victor, Rense, Rong, Febriyani, Adithya, Ioana, Anja, David, Joana,Mohammadreza,JonathanandMihaela.ItwasgreattomeatsomanynicepeoplethatwantedtodosomethingforthePhD’satouruniversity.ItwasalsoverynicetobethetreasurerofP‐NUTforalongtimetogetherwithJuanCarlos.

Alsothankstomyfootballteammatesattheuniversity.Iwasapleasuretoplaywithyouandhavesometimetorelaxnexttothehardwork.Bijoy,Daniël,Aram,Jorrit,Maurice,Andrea,Duc,Stelian,Felix,andalltheotherplayerswhoIplayedwith,thanksforthefunandhopefullywecanstillplayacoupleofmatchestogether.

Daarnaastwil ik ook graag Pim,Kasper en Carla bedanken voor de leuke dingen diewesamenhebbengedaan,zoalshetsamennaardebioscoopenhetuitetengaan.Ookwashetaltijderggezelligommet jullie te lunchen.Mooidatwesamenopkondentrekkende laatste jaren.KasperenPim:ikhoopookditjaarnogbijjulliepromotiesaanwezigtezijnenCarla:hopelijkkrijgjijeenvasteaanstellingaandeUT.VerderwilikookinhetbijzonderPeterbedankenvoordeleuketijddiewehebbengehadinEnschede.Ikhebhetaltijdergleukgevondenomsamentesnookerenen squashenenvindheterg leukdat jemijnparanimfwil zijn.Also thanks tomyotherparanymphArshdeep. Itwasverynice tobeyour roommateand thatwe coulddo thefinal steps in our PhD’s together, such that we could help each other.We had nice chats inbetweenthedifficultworkofwritingthethesis.AfterIwasyourparanymph,IamveryhappyyouwillbemyparanymphaswellandIwishyouagreatfutureinresearch.

Ookwil ik graagmijnhuisgenoten en oud‐huisgenoten vanCafé Zottekopbedanken.Hetwas altijd erg leuk om ’s avondna hetwerk thuis te komen en gezelligmet jullie te eten endaarnasomseenspelletjetespelen.Ikhebmealtijdergthuisgevoeldenhoopdatditookvoorjulliegeldt.Ikdenkdatweeenheleleuksfeerhebbenbijonsthuiswaarvanikhoopdatditintoekomstookzomagblijven.

Graachwol ikekmynHeit enMem,mynbroerKlaas, susterMariekeenhar freonRinsetankjefoarharrenstipeenfertrouwendy’tsemyaltydjûnhawwe.YnmynbeginperioadeynEnschedehie ik itnetaltydmaklik,mar jimbinnealtydeftermysteanbliuwnwêrtroch ik tabeslútditberikkekoe.Itisekaltydwergesellichomwiekeinsbyjimthústekommen.EkwolikHylke, Jitske en Ids betankje. Ek by jim fiel ik my thús. Wiekeinen faak puzzelje oan decryptogramofjûnsinspultsjedwaan.WilensdepromoasjehaikeknochlesjûnenwieikfakerynFryslântefinen.Altydkoeikbybeidefamyljesterjochteenstoenderiteneninbedfoarmyklear.Itishielmoaiomsulkefamyljetehawwen.EndanasletstemynfrouWilly.Skat,dostiestaltydfoarmyklearenwatfûnikitmoaiomdybymyynEnschedetehawwen.Wehahielwatôfreizge,maritresultaatmeiderwêze:dodynmasterpsychosomatischefysioterapyeniknomynpromoasje.Dynleafdeenfertrouwenhamydêrtigebyholpenenikhoapjedatwysnelinplakjetegearrefinemeie.

Harmen

size‐selective analyte detection in an biosensor...size‐selective analyte detection in an...

Documents