Floris Tallieu plasmonics Broadband colloidal quantum dot LED for active Academiejaar 2011-2012 Faculteit Ingenieurswetenschappen en Architectuur Voorzitter: prof. dr. Isabel Van Driessche Vakgroep Anorganische en Fysische Chemie Voorzitter: prof. dr. ir. Daniël De Zutter Vakgroep Informatietechnologie Master in de ingenieurswetenschappen: toegepaste natuurkunde Masterproef ingediend tot het behalen van de academische graad van Begeleider: Pieter Geiregat Promotoren: prof. dr. ir. Dries Van Thourhout, prof. dr. Zeger Hens
126
Embed
Broadband colloidal quantum dot LED for active plasmonics ...lib.ugent.be/fulltxt/RUG01/002/007/097/RUG01-002007097_2013_000… · ration and resulting in multiple pads of about 1
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Floris Tallieu
plasmonicsBroadband colloidal quantum dot LED for active
Academiejaar 2011-2012Faculteit Ingenieurswetenschappen en Architectuur
Voorzitter: prof. dr. Isabel Van DriesscheVakgroep Anorganische en Fysische Chemie
Voorzitter: prof. dr. ir. Daniël De ZutterVakgroep Informatietechnologie
Master in de ingenieurswetenschappen: toegepaste natuurkundeMasterproef ingediend tot het behalen van de academische graad van
Begeleider: Pieter GeiregatPromotoren: prof. dr. ir. Dries Van Thourhout, prof. dr. Zeger Hens
Preface
Five years of education in engineering physics at Ghent University, five years of gathering intel-
lectual knowledge. This master thesis is a crown of hard labour, perseverance and a continuous
search towards optimal solutions. It is the practical realization of theoretical approaches and
experimental hunches. Those are the reasons why this thesis would not have been as it is now
without the help, insight and support from other people.
Most of my acknowledgment goes to my supervisor, Pieter Geiregat, who fulfilled his role beyond
my expectations. Next to his technical support and already acquired knowledge, he provided
help with practical issues and even prepared some samples.
Next, gratitude should be offered to the joint between the Photonics Research Group (which is
part of the Department of Information Technology INTEC) and the Physics and Chemistry of
Nanostructures Group (PCN). The cooperation allowed to access more resources, more knowl-
edge and more interaction with the researchers. For this and for their guidance, I would like to
thank both of my promotors, prof. dr. ir. Zeger Hens and prof. dr. ir. Dries Van Thourhout.
However, not only the help of the people to whom you are accountable has to be mentioned,
also they who daily work next to you and help you with smaller or bigger issues deserve a word
of acknowledgment. Therefore, I would first like to thank everybody in both research groups for
the pleasant time I had, especially Antti, Stijn and Marco for providing quantum dots, Parvathi,
Kasia, Abdoulghafar and Tangi for their help during measurements and depositions and Bram
for his interest and helpful insights in my work.
In addition, people outside those research groups were of invaluable help. Not in the least prof.
dr. Philippe Smet and prof. dr. Diederik Depla who provided their material at my disposal,
combined with their experience and insights. Special thanks goes to Kilian for performing nu-
merous ALD depositions, Aimi for his experience with admittance measurements, Olivier for
XRD spectra and SEM images and to Francis and Wouter who helped me with the smallest
problems in the S1.
Of course, writing a thesis cannot be done individually, support from the people closest to
you is of crucial importance. First, my brother Steffen deserves to be mentioned for his support
and interest in my thesis subject. Next, special thanks goes to my girlfriend Tilde, who sup-
i
Preface
ported me to deliver a solid piece of work in a pleasant environment.
Finally, I would like to thank my parents for making me who I am, raising me in my best interest
and supporting me in my decisions. Without them, this master thesis would never have seen
the light.
Floris Tallieu
May 2012
ii
Copyright statement
The author gives permission to make this master dissertation available for consultation and to
copy parts of this master dissertation for personal use.
In the case of any other use, the limitations of the copyright have to be respected, in particular
with regard to the obligation to state expressly the source when quoting results from this master
dissertation.
Floris Tallieu
May 2012
iii
Broadband colloidal quantum dotLED for active plasmonics
by Floris Tallieu
Master thesis submitted to obtain the academic degree of
Master in engineering: applied physics
Academic year 2011-2012
Promotors: prof. dr. ir. Zeger Hens, prof. dr. ir. Dries Van Thourhout
Supervisor: ir. Pieter Geiregat
Faculty of Engineering and Architecture
Ghent University
Department of Information Technology (INTEC)
President: prof. dr. ir. Daniel De Zutter
Department of Inorganic and Physical Chemistry
President: prof. dr. Isabel Van Driessche
Summary
In this work, two types of LEDs are designed, fabricated and characterized. Their structure is
based on a combination of colloidal nanocrystals and inorganic layers. The first method uses
different layers to improve the transport of carriers when a DC voltage is applied, while the
second operation mechanism consists of two insulating layers which, when an AC voltage is
applied, imply the emergence of an electric field over the quantum dots, resulting in transport
of carriers, allowing radiative recombination of electron-hole pairs.
The characteristics of the layers are investigated, mainly focusing on their resistivity, mobility,
surface morphology and influence on the luminescence of the quantum dots. Furthermore,
electroluminescence experiments have been performed on both types, which showed most success
for the AC stacks.
Integration in a plasmonic structure is shortly discussed by analyzing a dipole emittor in a
metal-dielectric-metal structure, while stacks, based on the AC operation mechanism, have been
Broadband colloidal quantum dot LED for active plasmonics
Floris Tallieu
Ghent University, Faculty of Engineering and Architecture,
Department of Information Technology, Department of Inorganic and Physical Chemistry
Abstract In this work, two types of LEDs are designed, fabricated and characterized. Their structure is based on a combi-nation of colloidal nanocrystals and inorganic layers. The first method is based on the transport of carriers with a DC voltageapplied, while the second operation mechanism is based on an AC voltage-driven system with insulating layers. Characteriza-tion of the layers and electroluminescence experiments have been performed. Integration in a plasmonic structure is shortlydiscussed by analyzing a dipole emittor in a metal-dielectric-metal structure, while experimental stacks have been integratedon silicon waveguides.
Integration of light-emitting devices in plasmonic structuresor on silicon is a promising field. The applications of inte-grated plasmonics are ranging from x-ray devices to sensing,light trapping and ICs, while the advantages of CMOS com-patible silicon with the advanced functionalities of photoniccomponents are already widely tested [1]. Indeed, applicationsof silicon in telecommunications, biosensing, etc. are alreadyknown for their passive functionalities (signal routing, filtering,. . . ), yet the active photonics such as on-chip light generationor detection proves to be more challenging, mainly due to theindirect band gap of silicon. Therefore, different light-emittingdevices with quantum dots have been developed. Their opera-tion mechanism, device structure and results are discussed andfollowed by a short discussion on their integration. The quan-tum dots are solution-based colloidal nanocrystals, obtainedthrough a wet chemicable synthesis procedure and tunableto emit from the UV to the mid-infrared with high quantumyield [2].
II. Operation mechanisms
Two types of devices have been produced, each based on adifferent operation mechanism.
A. Direct charge injection - DC
The first electroluminescence method is based on direct chargeinjection. A schematic of the stack is shown in fig. 1.
Figure 1: Schematic of the stack based on direct charge injectionto induce electron-hole pairs in the quantum dot layer.
The electron transport layer (ETL) and the hole transport layer(HTL) help improve the injection of carriers into the quantumdots by reducing the energy barrier between the electrodes andthe energy level of the QD energy bands. Therefore, the ETL ismostly n-type, since the majority of carriers are electrons, whilethe HTL is mostly p-type [3]. The stacks are deposited on anITO covered glass substrate, which also serves as cathode. Theanode is always aluminium, deposited by electron beam evapo-ration and resulting in multiple pads of about 1.8 mm×1.8 mm.
B. Field-driven ionization - AC
The AC stacks by field-driven ionization are built with insu-lating layers. A schematic is shown in fig. 2.
Figure 2: Schematic of the stack based on field-driven ionizationto induce electron-hole pairs in the quantum dot layer.
Here, an AC voltage up to 100 V peak-to-peak is applied be-tween the ITO and the aluminium, resulting in the emergenceof a high electric field due to the capacitance series, induced byboth insulating layers and the quantum dot layer. Indeed, ifthe voltage drop per quantum dot exceeds the band gap, elec-trons in the valence band will tunnel to the conduction bandof the neighboring dot, while holes will make the exact oppo-site movement. This induces the possibility of simultaneouslyhaving an electron in the conduction band and a hole in thevalence band of the same quantum dot. This might lead toradiative recombination and photon emission. Since electronswill propagate to one side and holes to the other, an inter-nal field will build up, opposing the externally applied field.This diminishes the voltage drop per QD (or equivalently thelocal field strength). When this drop is no longer sufficientlyhigh, carrier transport from CB to VB or vice versa does nolonger take place. However, when the polarity is changed, theidentical process will occur in the opposite direction, since theelectric field of course reversed polarity as well [4].
III. Device structures and optimal layer composition
The choice of the materials for the transport layers in the DCstacks and the insulating layers in the AC stacks is of coursecrucial. Since there is strived towards the use of inorganic com-pounds to avoid fast degradation, nickel oxide and copper oxidehave been investigated as HTL. The latter however showed nop-type properties when DC magnetron sputtered. Nickel oxidehas been deposited by the same deposition method with a fixedoxygen pressure to yield a resistivity of 0.07 − 0.8 Ωcm for a30 to 60 nm thin film. For the ETL, zinc oxide is used, or inquantum dots, or deposited by ALD. Although they seemedviable options from literature [5], most research went to ro-tatable DC magnetron sputtered ZnO:Al (2 wt% Al), sincemultiple parameters can be changed, including the argon andoxygen pressure, the target-substrate distance and the thick-ness of the layer. They have been chosen to yield a resistivityof 0.4 Ωcm for a thickness of 75 nm.The material choice for the insulating layers in the AC stackswas pretty straightforward, since deposition of alumina couldbe done by electron beam evaporation or by ALD. Investiga-tion of the quality of both layers clearly showed the inferiorityof the ones deposited by e-beam. Therefore, 40 nm layers havebeen deposited by atomic layer deposition, both for the topand bottom layer. Admittance measurements also yielded a
v
value of ∼ 7.5 for the relative permittivity which is in goodagreement with literature [6].
IV. Results
Next to the extensive characterization of the ZnO:Al layer, theeffect of adding this or an alumina layer on top of the quantumdots have been investigated. Therefore, emission spectra, life-time measurements, transmission spectra and excitation spec-tra have been measured, which showed that the decrease inintensity of the photoluminescence spectra is due to the trans-mission characteristic of the extra layers and extra channelsfor non-radiative recombination, which reduces the lifetime ofthe photons. For the alumina layer, photoluminescence is onlyconserved for core-shell quantum dots/rods. Current-voltagemeasurements on the total stacks were quite similar to liter-ature results, but their irreproducibility, probably due to theaccumulation of charges at interfaces and electrodes, are anindication why no electroluminescence has been measured forthe DC stacks.Admittance measurements on the AC stacks and their indi-vidual layers showed that the voltage drop per quantum dothad to be at least the band gap energy to show electrolumi-nescence, which is predicted by the model of field-driven ion-ization. Electroluminescence have been measured for core-shelldots/rods emitting in the visible and in the infrared. An ELand PL spectrum are shown in fig. 3 for a CdSe/CdS quantumdot layer emitting in the red part of the visible spectrum.
1.0
0.8
0.6
0.4
0.2
0.0
Nor
mal
ized
inte
nsity
[a.u
.]
700650600550500λ [nm]
PL EL
Figure 3: EL and PL spectrum for an AC stack with CdSe/CdSquantum dots emitting in the red part of the visible spec-trum.
The electroluminescence spectrum is in good agreement withthe PL spectrum, indicating the same emission principle. Toshow the operation mechanism, the intensity of the peak dur-ing one period of the applied block wave of 20 kHz is depictedin fig. 4. Indeed, the high electric field at the beginning of thepositive pulse induces a sharp intensity increase, followed by arather linear decrease due to the emerging internal field.
inte
nsity
[a.u
.]
50403020100t [µs]
-100
0
100
voltage [V]
Run 1 Run 2 Voltage source
Figure 4: Light intensity in function of time.
The same mechanism happens in the opposite direction whenthe negative block pulse is initiated. The peak at exactly oneperiod might be caused by the internal field itself, which isstrong enough to induce light emission when the polarity ischanged.
The power in function of the applied voltage (at a frequency of20 kHz) is depicted in fig. 5 for the same stack, together withits linear fit. The power also shows a linear increase with thefrequency in the 1 − 15 kHz range.
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Pow
er [µ
W]
959085807570Voltage (pk-pk) [V]
Data points Linear fit
Figure 5: Output power of the AC stacks with CdSe/CdS dots(70 nm QD layer) in function of the applied voltage.
A maximum power of 2 µW has been measured. Based on theintercept, the underlying theory of the operation mechanism isconfirmed.
V. Integration
A model where a quantum dot, modeled as a dipole source,emits in the dielectric layer of an MDM structure, has beenconstructed to analytically calculate the spontaneous emissionenhancement factor. The resonant peak in the spectral resultis caused by the resonance of the surface plasmon polariton, asexpected.Next to this plasmonic model, an integrated structure has beendesigned to couple the light into a waveguide. A conceptualschematic is shown in fig. 6.
Figure 6: Threedimensional schematic of the integrated structureon silicon waveguides.
To check the influence an AC voltage over the quantum dotlayer, passive loss measurements have been performed forchanging voltage and changing area of deposited dots on thewaveguide. On top of the silicon waveguide, a monolayer ofquantum dots was deposited, then a 50 nm ALD layer of alu-mina and an ITO layer of 20 nm for top contacting. The siliconwaveguide layer itself is used as the bottom contact. An aver-age value of 2008 dB/cm is obtained for the loss caused by thequantum dot layer.
VI. Conclusion and future plans
The electroluminescence of the bulk AC devices with quan-tum dots emitting in the visible and near-infrared region showsthat this path of light generation looks promising, further effortshould be put into its characterization and further integrationfor plasmonic and silicon photonics purposes. Important as-pects supporting the latter are its compatibility with standardprocessing technologies, as well as the versatility of the plat-form, e.g. the easy combination with classic resonators suchas nitride microdisks. Furthermore, the characterization of theZnO:Al layer can be used to conduct different modulation ex-periments.
[2] Y. Yin and A. P. Alivisatos. Colloidal nanocrystal synthesisand the organic-inorganic interface. Nature, 437(7059):664–670, September 2005.
[3] J. M. Caruge, J. E. Halpert, V. Wood, V. Bulovic, andM. G. Bawendi. Colloidal quantum-dot light-emittingdiodes with metal-oxide charge transport layers. NaturePhotonics, 2(4):247–250, March 2008.
[4] V. Wood, M. J. Panzer, D. Bozyigit, Y. Shirasaki,
I. Rousseau, S. Geyer, M. G. Bawendi, and V. Bulovic.Electroluminescence from nanoscale materials via field-driven ionization. Nano letters, 11(7):2927–32, July 2011.
[5] J. Heo, Z. Jiang, J. Xu, and P. Bhattacharya. Coherentand directional emission at 1.55 µm from PbSe colloidalquantum dot electroluminescent device on silicon. Opticsexpress, 19(27):26394–8, December 2011.
[6] M. D. Groner, F. H. Fabreguette, J. W. Elam, and S. M.George. Low-temperature Al2O3 atomic layer deposition.Thin Solid Films, (16):639–645, 2004.
vii
Broadband colloidal quantum dot LED for active plasmonics
Floris Tallieu
Ghent University, Faculty of Engineering and Architecture,Department of Information Technology, Department of Inorganic and Physical Chemistry
Abstract In dit werk worden twee types LED’s ontworpen, gefabriceerd en gekarakteriseerd. Hun structuur is gebaseerdop een combinatie van colloıdale nanokristallen en anorganische lagen. De eerste methode is gebaseerd op het transport vanladingsdragers door het aanleggen van een DC spanning, terwijl het tweede werkingsmechanisme gebaseerd is op een AC spanningmet isolerende lagen. Karakterisatie van de lagen en elektroluminescentie-experimenten werden uitgevoerd. De integratie ineen plasmonische structuur wordt kort besproken door analyse van een dipoolbron in een metaal-dielektricum-metaal-structuur,terwijl experimentele devices werden geıntegreerd op een silicium golfgeleider.
Integratie van lichtemitterende devices in plasmonische struc-turen of op silicium is een veelbelovend researchdomein. Detoepassingen van geıntegreerde plasmonische structuren reikenvan x-ray devices tot sensoren, lichtvangst en geıntegreerde cir-cuits, terwijl de voordelen van CMOS compatibel silicium metde geavanceerde functionaliteiten van fotonische componentenreeds uitgebreid getest zijn [1]. Immers, toepassingen met si-licium in telecommunicatie, biosensoren, etc. zijn reeds ge-kend voor hun passieve functies (signaalsturing, filtering, . . . ).Desalniettemin zijn de actieve fotonische toepassingen zoalson-chip lichtgeneratie of -detectie een grotere uitdaging, gro-tendeels door de indirecte bandkloof van silicium. Daarvoorwerden verschillende lichtemitterende devices ontwikkeld metkwantum dots. Hun werkingsmechanisme, devicestructuur enresultaten worden besproken en gevolgd door een korte dis-cussie over hun integratie. De kwantum dots zijn colloıdalenanokristallen, geproduceerd in oplossing door een natte che-mische synthese en afstembaar om te emitteren in het UV totin het midden-infrarood met een hoge quantum yield [2].
II. Werkingsmechanismes
Twee devicetypes zijn geproduceerd, elk van hen gebaseerd opeen verschillend werkingsmechanisme.
A. Directe ladingsinjectie - DC
De eerste elektroluminescentiemethode is gebaseerd op directeladingsinjectie. Een schema van het device is weergegeven infig. 1.
Figuur 1: Schema van het device gebaseerd op directe ladingsin-jectie om elektron-gat-paren te creeren in de QD laag.
De elektronentransportlaag (ETL) en de gatentransportlaag(HTL) helpen de injectie van ladingsdragers in de kwantumdots te verbeteren door de energiebarriere tussen de elektro-des en het energieniveau van de kwantum dots te verlagen.Om die reden is de ETL meestal n-type, aangezien de meer-derheidsladingsdragers elektronen zijn, terwijl de HTL meestalp-type is [3]. De lagen zijn gedeponeerd op een glassubstraatbedekt met ITO, die ook meteen als de kathode gebruikt wordt.De anode is altijd aluminium, gedeponeerd via het opdampenm.b.v. een elektronenbundel en resulterend in meerdere vier-kante devices van ongeveer 1.8 mm × 1.8 mm.
B. Veld-gedreven ionisatie - AC
De AC devices door veld-gedreven ionisatie zijn opgebouwdmet isolerende lagen. Een schematische voorstelling is weerge-geven in fig. 2.
Figuur 2: Schema van het device gebaseerd op veld-gedreven ioni-satie om elektron-gat-paren te creeren in de QD laag.
Hier wordt een AC spanning tot 100 V peak-to-peak aange-legd tussen de ITO-laag en het aluminium, wat een hoog elek-trisch veld opwekt door de serie condensatoren, gevormd doorbeide isolerende lagen en de kwantum dot laag. Immers, als despanningsval per kwantum dot groter wordt dan de band gap,zullen de elektronen in de valentieband naar de conductiebandvan de naburige dot tunnelen, terwijl de gaten de exact tegen-gestelde beweging zullen maken. Dit creeert de mogelijkheidom terzelfdertijd een elektron in de conductieband en een gatin de valentieband van dezelfde kwantum dot te krijgen. Ditkan vervolgens tot radiatieve recombinatie leiden en bijhorendeemissie van een foton. Aangezien elektronen zullen propagerennaar de ene kant en gaten naar de andere, zal een intern veld op-gebouwd worden, tegengesteld aan het extern aangelegde veld.Dit verlaagt de spanningsval per kwantum dot (of equivalentde sterkte van het lokale veld). Wanneer deze spanningsvalniet langer hoog genoeg is, wordt het ladingstransport van deconductieband naar de valentieband of vice versa onderdrukt.Wanneer de polariteit wordt omgewisseld, vindt het identiekeproces plaats in de tegengestelde richting, aangezien het elek-trisch veld natuurlijk ook van polariteit veranderd is [4].
III. Devicestructuren en optimale laagsamenstelling
De keuze van de materialen voor de transportlagen in de DCdevices en de isolerende lagen in de AC devices is natuur-lijk cruciaal. Omdat er gestreefd wordt naar het gebruikvan anorganische componenten om snelle degradatie te ver-mijden, werden vooral nikkeloxide en koperoxide onderzochtals HTL. Deze laatste bleek echter geen p-type eigenschappente vertonen wanneer ze gedeponeerd werd via DC magnetronsputtering. Nikkeloxide werd op dezelfde manier gedeponeerdmet een vaste zuurstofdruk, resulterend in een resistiviteit van0.07 − 0.8 Ωcm voor een 30 tot 60 nm dunne film. Voor deETL werd zinkoxide gebruikt, ofwel als kwantum dots, ofwelgedeponeerd via ALD. Hoewel deze goede opties bleken uit deliteratuur [5], werd de meeste research gedaan naar ZnO:Al(2 wt% Al), gedeponeerd via roterende DC magnetron sputte-ring, aangezien veel parameters gewijzigd kunnen worden, o.a.de argon- en zuurstofdruk, de target-substraatafstand en dedikte van de laag. Deze werden zo gekozen om een resistiviteitte bekomen van 0.4 Ωcm voor een dikte van 75 nm.De materiaalkeuze voor de isolerende lagen van de AC deviceswas behoorlijk eenvoudig, aangezien de depositie van aluminakon gedaan worden door zowel opdamping met een elektronen-bundel of atomaire laag depositie. Analyseren van de kwaliteitvan de beide lagen toonde echter duidelijk de minderwaardig-heid van degene gedeponeerd via de elektronenbundel. Daarom
viii
werden 40 nm lagen gedeponeerd via ALD, zowel voor de bo-venste als de onderste laag. Admittantiemetingen leverden eenwaarde van ∼ 7.5 op voor de relatieve permittiviteit, wat ingoede overeenkomst is met de literatuur [6].
IV. Resultaten
Naast de uitgebreide karakterisatie van de ZnO:Al-laag, werdhet effect van het deponeren van deze of een aluminalaag bo-venop de kwantum dots onderzocht. Daarvoor werden emissie-,transmissie-, excitatiespectra en de levensduur opgemeten,waaruit werd besloten dat de afname van de intensiteit in defotoluminescentiespectra te wijten is aan de transmissiekarak-teristiek van de extra lagen en aan extra kanalen voor niet-radiatieve recombinatie, wat de levensduur van de fotonen sterkinkort. Stroom-spanningskarakteristieken op de complete de-vices waren in goede overeenkomst met literatuurresultaten,maar het gebrek aan reproduceerbaarheid, waarschijnlijk doorde opstapeling van ladingen aan interfaces en elektroden, is eenaanwijzing waarom geen elektroluminescentie opgemeten werdvoor de DC devices.Admittantiemetingen op de AC devices en hun individuele la-gen toonden aan dat de spanningsval per kwantum dot min-stens de bandkloofenergie diende te zijn om elektrolumines-centie te bekomen, wat ook voorspeld is door het model vande veld-gedreven ionisatie. Elektroluminescentie werd geme-ten voor core-shell dots/rods, emitterend in het visuele en IRspectrum. Een elektro- en fotoluminescentiespectrum wordtweergegeven in fig. 3 voor een kwantum dot laag, die emitteertin het rode gedeelte van het zichtbare spectrum.
1.0
0.8
0.6
0.4
0.2
0.0
Nor
mal
ized
inte
nsity
[a.u
.]
700650600550500λ [nm]
PL EL
Figuur 3: EL en PL spectra voor een AC device met kwantumdots die emitteren in het rode gedeelte van het zichtbarespectrum.
Het elektroluminescentiespectrum is in goede overeenkomstmet het PL spectrum, wat aantoont dat hetzelfde emissieprin-cipe plaatsvindt. Om het werkingsmechanisme aan te tonen,wordt de intensiteit van de piek gedurende een periode van deaangelegde blokgolf van 20 kHz weergegeven in fig. 4. Zoalsverwacht veroorzaakt het hoge elektrisch veld aan het beginvan de positieve puls een scherpe intensiteitstoename, gevolgddoor een eerder lineaire afname door het opkomende interneveld.
inte
nsity
[a.u
.]
50403020100t [µs]
-100
0
100
voltage [V]
Run 1 Run 2 Voltage source
Figuur 4: Emissie-intensiteit in functie van de tijd.
Hetzelfde mechanisme gebeurt in de tegengestelde richtingwanneer de negatieve blokpuls bereikt wordt. De piek zicht-baar op exact een periode is mogelijks veroorzaakt door hetinterne veld zelf, maar hier werd geen bevestiging voor gevon-den.Het vermogen in functie van de aangelegde spanning (bij een
frequentie van 20 kHz) is weergegeven in fig. 5 voor dezelfdestack, samen met een lineaire fit.
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Pow
er [µ
W]
959085807570Voltage (pk-pk) [V]
Data points Linear fit
Figuur 5: Lichtvermogen van de AC stacks met CdSe/CdS dots(70 nm QD laag) in functie van de aangelegde spanning.
Het maximaal opgemeten vermogen bedraagt 2 µW. Op basisvan het intercept, werd de onderliggende theorie van het wer-kingsmechanisme bevestigd. Het lichtvermogen vertoont ookeen lineaire toename bij toenemende frequenties in een bereiktussen 1 en 15 kHz.
V. Integratie
Er werd een model gemaakt waar een kwantum dot, gemodel-leerd als een dipoolbron, emitteert in de dielektrische laag vaneen MDM structuur, om analytisch de spontane emissie en-hancement factor te berekenen. Dit levert, voor een voldoendedunne dielektrische laag, een resonante piek in het spectraleresultaat, veroorzaakt door de gap SPP mode, zoals verwacht.Naast dit plasmonisch model werd ook een geıntegreerde struc-tuur ontwikkeld om licht te koppelen in een silicium golfgelei-der. Een conceptueel schema wordt weergegeven in fig. 6.
Figuur 6: Driedimensionaal schema van de geıntegreerde structuurop een silicium golfgeleider.
Om de invloed van een AC spanning over de kwantum dotlaag te onderzoeken, werden passieve verliesmetingen uitge-voerd voor een veranderende spanning en varierende opper-vlakte van gedeponeerde dots op de golfgeleider. Bovenop desilicium golfgeleider werden een monolaag van kwantum dotsgedeponeerd, vervolgens een 50 nm ALD alumina laag en eenITO-laag van 20 nm, die als bovencontact fungeert. De silici-umlaag zelf wordt gebruikt als het andere contact. Een gemid-delde waarde van 2008 dB/cm werd bekomen voor het verliesveroorzaakt door de kwantum dot laag, wanneer geen spanningwerd aangelegd.
VI. Conclusie en toekomstplannen
De elektroluminescentie van bulk AC devices met kwantumdots die emitteren in het zichtbare en nabije infrarood spec-trum toont aan dat dit pad van lichtgeneratie veelbelovendlijkt, waardoor inspanningen zouden moeten gedaan wordenom verdere karakterisatie en integratie in plasmonische struc-turen en op silicium golfgeleiders door te voeren. De compa-tibiliteit met standaard processingtechnologieen en de veelzij-digheid van het platform, zoals de eenvoudige combinatie metklassieke resonatoren zoals nitride microdisks, zijn enkele be-langrijke aspecten die dit verder ondersteunen. Bovendien kande karakterisatie van de ZnO:Al laag gebruikt worden om ver-schillende modulatie-experimenten uit te voeren.
[2] Y. Yin and A. P. Alivisatos. Colloidal nanocrystal synthesisand the organic-inorganic interface. Nature, 437(7059):664–670, September 2005.
[3] J. M. Caruge, J. E. Halpert, V. Wood, V. Bulovic, andM. G. Bawendi. Colloidal quantum-dot light-emitting dio-des with metal-oxide charge transport layers. Nature Pho-tonics, 2(4):247–250, March 2008.
[4] V. Wood, M. J. Panzer, D. Bozyigit, Y. Shirasaki, I. Rous-seau, S. Geyer, M. G. Bawendi, and V. Bulovic. Electrolu-minescence from nanoscale materials via field-driven ioni-zation. Nano letters, 11(7):2927–32, July 2011.
[5] J. Heo, Z. Jiang, J. Xu, and P. Bhattacharya. Coherentand directional emission at 1.55 µm from PbSe colloidalquantum dot electroluminescent device on silicon. Opticsexpress, 19(27):26394–8, December 2011.
[6] M. D. Groner, F. H. Fabreguette, J. W. Elam, and S. M.George. Low-temperature Al2O3 atomic layer deposition.Thin Solid Films, (16):639–645, 2004.
6.1 Summary of both operation mechanisms, their advantages and disadvantages. . . 76
xx
List of abbreviationsAlq3
AZO
DEZ
DOS
ETL
EQE
HDA
HOMO
HTL
ITO
IV
LED
MDM
NA
NC
ODE
OLED
OSA
PL
QD
QR
QY
SC
SEM
SILAR
SPP
TDPA
TM
TOP
TPD
ZTO
Tris(8-hydroxyquinolinato aluminium)
Aluminium-doped Zinc Oxide
Diethyl Zinc (Zn(CH2CH3)2)
Density Of States
Electron Transport Layer
External Quantum Efficiency
Hydroxydopamine
Highest Occupied Molecular Orbital
Hole Transport Layer
Indium Tin Oxide
Current-Voltage
Light Emitting Diode
Metal-Dielectric-Metal
Numerical Aperture
Nanocrystal
1-octadecene
Organic Light Emitting Diode
Optical Spectrum Analyzer
Photoluminescence
Quantum Dot
Quantum Rod
Quantum Yield
Spincoated
Scanning Electron Microscopy
Sequential Ion Layer Addition and Reaction
Surface Plasmon Polariton
Thiodipropionic Acid
Transverse Magnetic
Trioctylphosphine
N,N′-bis(3-methylphenyl)-N,N′-diphenylbenzidine
Zinc Tin Oxide
xxi
Chapter 1
Introduction
1.1 Context
Quantum dots are a promising research area in a lot of domains. Solar cells, photovoltaics,
medical imaging and disease detection, anti-counterfeiting capabilities, even counter-espionage
and defense applications are current areas of investigation [1–8]. Mainly however, quantum dots
might have its most promising applications in light emitting diodes. They can be tuned to emit
in a wide range of wavelengths, ranging from blue to infrared, meaning that they can be used
for display applications or even in telecommunication. To show the possibilities of quantum dot
LEDs, a comparison of the color gamut is made with classical liquid crystal displays. This is
shown in fig. 1.1, where we clearly see that the spectral purity of quantum dots is higher than
a standard HDTV color triangle.
Figure 1.1: CIE chromaticity diagram for quantum dots and a standard HDTV color triangle. [9]
Of course, air-stable components are necessary to be able to produce this in an inexpensive way.
Therefore, in this thesis, we strived to work with inorganic components as surrounding layers
for the QDs, since they show significantly slower degradation in time, caused by the negative
1
Chapter 1. Introduction
influence of air and moist on organic compounds.
The second part of the thesis title refers to plasmonics. This term comprises the optical prop-
erties of metal structures at the nanoscale. More specific, nowadays lots of applications are
under investigation involving the use of surface plasmons. Some examples are shown in fig. 1.2,
while the theoretical approach is explained in chapter 7. As should be clear, the applications
(a) Creation of soft
x-rays by a hot spot.
(b) Efficient sensing due
to resonance shifting mea-
surement.
(c) Enhanced light trap-
ping in a solar cell.
(d) Plasmonic integrated cir-
cuit.
Figure 1.2: Examples of applied plasmonics. [10]
of plasmonic structures are very wide, ranging from the creation of soft x-rays1 by a hot spot2
to sensing, light trapping and integrated circuits [10, 11]. Furthermore, research projects have
already been initiated to use the silicon-plasmonic platform for optical interconnection [12].
Here, an active plasmonic application will be discussed by using the quantum dot LED structure
in between two metal plates, resulting in an optical source in an MDM structure. This might
link an electronic circuit with the photonic one in an easy and inexpensive way.
Finally, integration of the light source on silicon remains a field which cannot be ignored, because
of its compatibility with standard processing techniques, as well as the versatility of the plat-
form, e.g. the easy combination with classic resonators such as nitride microdisks. That’s why
in this thesis, one type of the LEDs will be integrated on a silicon waveguide and subsequently
be analyzed.
1.2 Motivation
Choosing a master thesis was one of the most difficult things to do as student civil engineering
in applied physics. Not only you want it to be fun, you also want it to be rewarding for you and
everybody around you. Therefore I looked first of all to my own interests, which lie in photonic
applications and in quantum mechanics. Of course, finding something which combines both is not
easy. However, intrigued by the word quantum dot of which I barely knew anything, this master
thesis appealed to me because of multiple reasons, in first instance due to its variety of work,
1X-rays with an energy between 12 and 120 keV.2Spot where the intensity of an incoming light beam can be concentrated by more than four orders of magnitude.
2
Chapter 1. Introduction
from practical experimental to theoretical simulations. Furthermore it attracted me because it
is a collaboration between two research groups, both important in the science community. All
of these mentioned elements made me consider this thesis as a challenge, as I preferred it.
1.3 Goals
The title can clearly be split into two parts. First of all, a working quantum dot LED should be
made, where attention should be put to its air-stable operation. Therefore, different inorganic
materials should be characterized in order to make layers with optimal electronic and optical
properties. Also emphasis should be put on the broadband operation, where the same LED
stack could be used with a wide wavelength range of different quantum dots.
Next to the quantum dot LED, its integration into plasmonic materials should be optimized.
To achieve this, simulations considering the quantum dots in the LED stack as optical source
of photons propagating along metal-dielectric(-metal) structures are done. Finally, one type of
the LEDs has been integrated on a silicon waveguide and associated measurements have been
performed.
1.4 Thesis report structure
In chapter 2, an extensive introduction of colloidal nanocrystals is given, where the theoretical
and the practical part are explained, as well as different types of nanocrystals. In chapter 3,
a closer look is taken to the composition of the quantum dot LED, elaborating on the general
structure, on all different layers individually and on the operation mechanisms. Next, all used
deposition, measurement and characterization methods are discussed in chapter 4, ranging from
IV measurements to SEM pictures. Based on these, characteristics of all layers are determined
and optimized for the LED stack in the next chapter. Chapter 6 shows various kinds of mea-
surement results on this total stack. Finally, in chapter 7, we arrive, on the one hand, at the
integration into a plasmonic structure, where an MDM with optical source is analyzed. On the
other hand, loss measurements of an integrated structure on a silicon waveguide are performed.
3
Chapter 2
Colloidal nanocrystal quantum dots
To be able to go into detail on colloidal nanocrystal quantum dots, all phrase parts need a
profound understanding. Therefore, an elaborate explanation about quantum dots is given,
followed by the limitation to colloidal nanocrystals.
2.1 Quantum dots
2.1.1 Definition
According to the Encyclopedia of Laser Physics and Technology, quantum dots are defined as
microscopic structures confining charge carriers in three dimensions. This charge confinement
is in fact a quantum effect, where the part ‘dot’ points out that this happens in all three spatial
dimensions. A first appearance of the term occurred in 1988 by Reed [13]. However, it was
already in the early 80’s that materials have been discovered which behave according to the
above definition [14].
2.1.2 Theoretical approach
As explained in the previous subsubsection, three-dimensional spatial confinement occurs. This
reflects itself in the dimensions of the structure, which are in every direction of the order of
nanometers. This physical range is of importance because it has the same order of magnitude as
the exciton Bohr radius, which indicates the most probable distance between the electron and
the hole for an electron-hole pair in a bulk material. Conclusively, in the length scale regime
below some tens of nanometers, strong confinement occurs which quantum mechanically only
yields discrete energy levels inside the quantum well [15]. A pictorial representation of what
happens with the energy levels is shown in fig. 2.1.
This can quantum mechanically be explained based on the Schrodinger equation and the Pauli
exclusion principle. The time-independent Schrodinger equation for the potential well is defined
4
Chapter 2. Colloidal nanocrystal quantum dots
(a) Bulk semiconductor. (b) Quantum dot.
Figure 2.1: Comparison of the band diagram of a bulk semiconductor and a quantum dot.
as:
− ~2
2m∇2Ψ(r) + V (r)Ψ(r) = EΨ(r) (2.1)
where m denotes the effective mass of the particle and ∇2 is the Laplace operator.
For an infinite square potential well in one dimension, this results in easy solutions for the bound
energy states:
En = n2h2
8mL2(2.2)
This indeed shows that the smaller the well, the further the energy levels separate.
Since an infinite square potential well in one dimension is not a good representation for the
zero-dimensional case of the spherical quantum dot, a better approximation can be made by
considering the bound energy states of a sphere with infinite square potential well. These are
given analytically (even for the finite case) by [16]:
Ekl =h2
8πm
(aklR
)2(2.3)
In this equation, akl is the kth zero of the lth spherical Bessel function of the first kind. It
should be clear that the same inverse squared dependence on the well radius is found as in eq.
(2.2).
To refer back to the original depiction of the quantum dot band diagram in fig. 2.1(b), we can
easily calculate the new band gap, taking into account the original band gap Eg, the effective
masses of the holes mh and electrons me and the fact that the first zero to be considered is a00:
Eg,QD = Eg +h2
8π
(a00R
)2( 1
mh+
1
me
)= Eg +
h2
8πµ
(a00R
)2(2.4)
As expected, the band gap becomes larger due to the quantum confinement. The change also
scales with 1/R2, showing that smaller particles shift the absorption to higher energies (or
shorter wavelengths). This simplified expression can be expanded with two terms [17]:
A negative term scaling with 1/R which includes the Coulombic electron-hole interaction,
A position-dependent solvation energy term, arising from dielectric screening.
5
Chapter 2. Colloidal nanocrystal quantum dots
However, these additional terms do not undo the increasing band gap for decreasing diameter.
A last important subject in the discussion of the quantum confinement of quantum dots is
the density of states of zero-dimensional systems. To illustrate what happens, in table 2.1, a
graphical depiction and corresponding expressions of the DOS are presented when a subsequent
confinement is performed from 3D to 0D. All dispersion relations are assumed parabolic. As
Dim Image DOS Graph
3 12π2
(2m∗
~2)3/2√
E − Ec
2∑n
m∗
π~2H(E − En)
1∑n
1π~
√m∗
2(E−En)
0∑n
2δ(E − En)
Table 2.1: Density of states of quantum confined semiconductors with a parabolic dispersion relation.
predicted in the discretization of the energy levels, only delta peaks arise in the 0D DOS. When
measured in reality, they are peaks width a certain width, caused by both homogeneous and
inhomogeneous broadening.
2.2 Colloidal nanocrystals
Nanocrystals are defined as a portion of matter where at least in one dimension the length scale
is smaller than 100 nm and the material is monocrystalline [18]. The creation of nanoscale
particles can be done in several ways: epitaxial, sputtering, ion implantation, precipitation in
molten glasses, chemical synthesis. . . [15]. The latter can be divided into different methods,
such as reverse micelles technique, sonochemical synthesis and colloidal synthesis [19]. Since in
this thesis only colloidal nanocrystals are used, an elaborate explanation of their synthesis and
the different types are given.
6
Chapter 2. Colloidal nanocrystal quantum dots
2.2.1 Synthesis of colloidal nanocrystals
The creation of colloidal nanocrystals is a bottom-up method, where the reaction of the precur-
sor(s) in a solution gives rise to the nucleation of small particles, which grow bigger the longer
the reaction is maintained. The nucleation phase was first described by La Mer and Dinegar,
who succeeded in synthesizing monodispersed sulphur colloids [20]. The growth phase, which
in fact consists of a rapid size increase and finally a slower process, known as Ostwald ripening,
where the larger particles grow and the smaller particles dissolve back to monomers.
2.2.1.1 Nucleation phase
As stated before, when the concentration of the precursor, which is injected in a hot solvent as
shown in fig. 2.2, reaches a sufficiently high level called the nucleation threshold, super-saturation
occurs. This results in the decomposition of the precursor in monomers and subsequent agglom-
Figure 2.2: Injection method of the precursor in a hot solvent for colloidal nanocrystal synthesis. [15]
eration of the latter. If no more precursor material is added to the solvent, the concentration
level will return below the nucleation threshold and the nucleation process will be put on hold.
Thermodynamically, the formation of the nanocrystal particles is a consequence of overcoming
an energy barrier, as depicted in fig. 2.3. The free energy difference between the solution and
the (n-monomer) nanocrystal is given by:
∆G = n(µc − µs) + 4πr2γ (2.5)
where µs and µc are the chemical potentials of the solution and the nanocrystal (per monomer)
respectively, r the particle radius and γ the surface tension. This formula can be rewritten and
approximated in function of the particle radius as single variable. The surface term dominates
the negative first term until the critical radius rc is reached as shown in the graph.
7
Chapter 2. Colloidal nanocrystal quantum dots
Figure 2.3: Energy barrier of the nucleation reaction. [21]
2.2.1.2 Growth phase
After nucleation, it is energetically favorable for the aggregates to grow. When there is a large
excess of monomers, the growth rate is determined by the rate of the monomer-surface reaction
and only affects the average radius, not the size distribution. This fast process is followed
by a diffusion-limited growth where the transport of monomers becomes the limited factor [21].
When the monomer supply is completely depleted, Ostwald ripening occurs. This phenomenon is
characterized by the detachment of monomers from the smaller particles, which are subsequently
used for the growth of the larger ones. A schematic representation of the equilibrium solubility
of a particle as a function of the particle radius is depicted in fig. 2.4, which shows that particles
smaller than the mean particle size will dissolve, whereas the particles larger than the mean
particle size will grow.
Figure 2.4: Equilibrium solubility of a particle as a function of the particle radius. [22]
Size dispersion is not easy to describe but is mainly determined by three parameters: the
monomer concentration, the surface tension and the temperature. Controlling these is not
evident, but literature and experience yield adequate results [21,23–26].
8
Chapter 2. Colloidal nanocrystal quantum dots
2.2.2 Types of colloidal nanocrystals
The basic principles of nucleation and growth of colloidal nanocrystals apply for all of them, yet
from thermodynamic point of view, a solution of nanoparticles is not a stable state. They have
to be stabilized by ligands, or by charge. Here, a small overview of attainable nanoparticles is
given.
The first syntheses of nanocrystals were performed in a nitrogen environment, and were mainly
cadmium and lead chalcogenides (sulfides, selenides and tellurides). Most of the syntheses are
already described by standardized methods including multiple parameters, such as size, size
dispersion, PL quantum yield, etc. In table 2.2, a small overview is given of the most common
chalcogenide nanocrystals, including alloys, which are cadmium with different ratios of selenium
and sulphur ODE precursors.
Material Size [nm] Size dispersion [%] PL QY [%] Ligands
CdSe 2− 4 6− 8 1− 5 Stearic acid & HDA
CdSe (blue) 1.6− 2.2 1− 6 1− 5 Stearic acid & HDA
CdS 6 5 Trap emission Fatty acid
CdTe 3− 11 4− 8 Unknown TDPA
Alloys 2.4− 3 5 10 Fatty acid
PbSe 3− 8 5− 9 40 Oleic acid
PbS 3− 10 5− 14 20− 90 Oleic acid
Table 2.2: Overview of the properties of chalcogenide nanocrystals.
The PL quantum yield (on which will be elaborated in the next section) can often be improved,
for example by adding a ZnS layer, by ligand exchange or by changing the core/shell ratios.
Three types of adapted syntheses deserve special attention:
Cation exchange reaction: Outer layers of the core are replaced by an other chalcogenide
(e.g. CdSe):
– PbSe: PL QY: 25− 40%
– PbS: PL QY: 10− 90%
SILAR: Layer by layer addition: PL QY: 1− 40%
Flash: Flush at lower temperatures, adding TOP-S and CdSe seeds at higher temperature:
PL QY: up to 60%
Not only quantum dots can be created in chemical synthesis, also rods, such as CdSe and
CdSe/CdS rods which have cylinder radii from 2 to 20 nm and a length smaller than 100 nm.
9
Chapter 2. Colloidal nanocrystal quantum dots
They can show even higher photoluminescence quantum yield than normal dots.
A last important synthesis of nanocrystals is the creation of ZnO quantum dots, which are not
stabilized by any ligands, but by charge. While the London-interaction between the colloids is
usually lowered by the adding of capping molecules, they can also be countered by the Coulomb
repulsion if the colloids carry an electric charge [22,27].
2.3 Optical and electrical properties of colloidal quantum dots
2.3.1 Light absorption and emission
As should be clear up to now, both the absorption and the emission of photons of the quantum
dots are strongly dependent on its size and its bulk band gap. This can be easily seen if we
look at an example of an absorbance spectrum of different dispersed quantum dots. For some
cadmium and lead chalcogenides, this is shown in fig. 2.5.
Figure 2.5: Absorbance spectra of cadmium and lead chalcogenides having different sizes. [27]
Some important conclusions can be drawn from these spectra:
Since they are all different semiconductors, they have a different band gap energy, resulting
in shifted spectra accordingly.
For example, CdS has a bulk band gap of 2.42 eV at 300 K, while CdSe has a band gap of
1.74 eV [28,29]. CdS will thus have absorption spectra at shorter wavelengths than CdSe.
The presence of peaks in each of the spectra refers to the discretization of the energy levels,
as was also clear in the density of states. The first absorption peak (agreeing with the light
with the longest wavelength/smallest energy to be absorbed) is of course most visible. On
10
Chapter 2. Colloidal nanocrystal quantum dots
the broadening of the peaks is elaborated further in this subsection.
Absorption spectroscopy also yields information about the size dispersion and the concen-
tration.
The size dispersion is easily measured by fitting a Gaussian curve to the first peak, while the
calculation of the concentration of the quantum dots is based on following formula:
I
I0= 100.0277d
3c0l (2.6)
where II0
denotes the fraction of the transmitted light measured at the peak, d the diameter of
the particles (immediately determined by the peak position), c0 the unknown concentration and
l the propagation length of the light through the cuvette (1 cm) [30].
Since we are mainly interested in the emission rather than the absorption, a plot of both spectra
might yield more information about the band diagram. An example for a CdSe quantum dot
is given in fig. 2.6. First of all, we see that the first absorption peak does not coincide with
Figure 2.6: Absorption (solid line) and photoluminescence spectrum (dashed line) for a CdSe quantum
dot. [31]
the emission peak. This is due to the splitting of the highest valence band, as depicted in fig.
2.7, caused by electron-hole exchange interaction [32]. In the absorption spectrum, only the
transition from the lowest one of both states is visible, since this one has the highest oscillator
strength. The energy transfer during absorption will thus be bigger than when the electron
relaxes to the highest energy state in the valence band. This Stokes shift results in a red-shift
of the PL spectrum.
Second, multiple reasons can be cited to cause the broadening of the peaks. Mainly, since the
spectra are measured for all of the illuminated dots (in solution or deposited on a substrate), a
broadening of the PL peak and the absorption peak occurs due to the size dispersion. Moreover,
coupling between phonons and electrons will cause an additional broadening. The occurence of
11
Chapter 2. Colloidal nanocrystal quantum dots
Figure 2.7: Different absorption and emission wavelength due to valence band splitting.
the absorption peaks at higher energy levels are less visible, because the energy eigenvalues de-
rived from the Schrodinger equation get closer together, resulting in a smaller energy separation
between the discretized levels, ending up in a continuum.
2.3.2 Radiative and non-radiative recombination
As we know, absorption of a photon results in the creation of an electron-hole pair where the
electron gets excited to the conduction band. There are however multiple ways of recombining.
They can be divided into two categories, radiative and non-radiative. The former is of course
discussed in the previous subsection, because it describes the phenomenon of emitting a photon
when recombination occurs. A schematical depiction is shown in fig. 2.8(a).
When non-radiative recombination takes place on the other hand, the energy is not transferred to
an emitting photon, but consumed in other ways. In quantum dots, the relative influence of the
different processes remains undetermined. It is however known that Auger recombination has a
significant contribution [32–34]. During this process, the relaxation energy of the recombining
Figure 6.15: Capacitance-frequency measurements of the total AC stack.
where dtop is assumed to be 0, since we do not expect a separate ALD alumina layer on top
of the 70 nm quantum dot layer, based on the SEM images in the previous chapter. The area
is approximated to be 3.5 mm2 and the total capacitance averaged to 1.52 nF. The resulting
capacitance over the QD layer is then calculated to be 2 nF. Since a peak-to-peak voltage of
about 72 V (so an amplitude of 36 V) has to be applied to measure light output (see further),
the voltage drop over the total quantum rod layer can be calculated based on eq. (3.1) and (3.3)
and yields:
VQD = 6.95 V (6.5)
The dots have a diameter of 14.1 nm, which implies 3 to 5 monolayers. The resulting voltage
drop per quantum dot can thus be determined between 1.4 and 2.3 V, which, based on the
electric field model, in good agreement with the energy band gap of the dots, 2.02 eV.
6.4.2 Extended model calculations
However, previous calculations cannot explain the capacitance change when the quantum rod
layer is spincoated once or twice. Based on the capacitance density-frequency plots of the
alumina layers and the total stack in fig. 6.16 and the conductance-frequency plots of the ITO-
Al2O3-Al stack in fig. 6.17, an extension of the simple model is made by considering extra
resistors, as already mentioned in 3.2. Fig. 6.16 also confirms the uniformity of the layers, since
the capacitance densities of two different pads are almost perfectly coinciding.
As is clearly visible, the conductance G of the oxide layer is not constant, but shows a dependence
on the frequency. Therefore, a power fit has been performed in the frequency range of interest
(5 − 100 kHz). This yields a power dependence of 1.94 on the frequency. Based on this result,
a first electrical circuit proposal for the alumina layers is depicted in fig. 6.18.
66
Chapter 6. Stack results
2
3
4
56
1
2
3
4
56
C [n
F/m
m²]
2 3 4 5 6 7 8
104
2 3 4 5 6 7 8
105
2 3 4 5 6 7 8
106
f [Hz]
Bottom layer 1 Bottom layer 2 Top layer 1 Top layer 2 Stack 1x red QDs 1 Stack 1x red QDs 2 Stack 2x red QDs 1 Stack 2x red QDs 2
Figure 6.16: Capacitance density-frequency plot of alumina layers and total stack, where the QD layer
is spincoated once or twice.
10-8
10-7
10-6
10-5
10-4
10-3
10-2
G [S
]
104
105
106
f [Hz]
Original data Power fit Model fit
Figure 6.17: Conductance-frequency plot and fitting curves of the ITO-Al2O3-Al, where the alumina
layer is the bottom oxide layer of the total stack.
Figure 6.18: Electrical circuit proposal for the ITO-Al2O3-Al stack.
Indeed, the admittance expression looks like:
Y = G+ jωCp (6.6)
=1 + ω2R2
2
(1 + R1
R2
)C2
R1(1 + ω2R22C
2)+ jω
C
1 + ω2R22C
2(6.7)
where Cp denotes the result of the Cf-measurements as plotted in all appropriate graphs. Indeed,
if ω2R22C
2 remains far smaller than 1 (for the frequency range of interest) and R1 R2, the
67
Chapter 6. Stack results
quadratic dependence is present, while the capacitance remains constant:G ≈1R1
+ ω2R2C2
Cp ≈ C(6.8)
The model fit to the conductance curve, together with the already obtained value of C, yields
the values: R1 = 3.32 GΩ
R2 = 36.6 Ω
C = 6.5 nF
(6.9)
showing that the parallel path can be discarded and the conditions for the model are fulfilled.
The original data, the power and model fit of G are shown in fig. 6.17. Analogous measurements
have been performed for the top layer, resulting in similar values for the parameters.
Now that the oxides are modeled, focus is put on the quantum dot/rod layer. Therefore, we
calculate its admittance by subtracting the admittances of the oxides from the total stack.
Similar to what is done in the simplified case, the top oxide admittance is discarded, since we
assume, based on the SEM images that no separate top alumina layer is grown. If we assume
that the imaginary part of the quantum dot layer admittance consists of a similar term as the
oxide layers, we obtain the capacitance of the quantum dot layer as in fig. 6.19.
1
2
3
4
5
6
789
10
C [n
F]
2 3 4 5 6 7 8
104
2 3 4 5 6 7 8
105
2 3 4 5 6 7 8
106
f [Hz]
1xSC 2xSC
Figure 6.19: Capacitance-frequency plot of the quantum dot layer (spincoated once or twice), based on
the extended model.
Contrary to what was obtained in the idealized model, the capacitance is no longer equal. The
capacitance of the thicker layer is about two-thirds of the one of the thinner layer, which is still
a deviation of the expected halving. Two possible deficiencies of the model can explain this:
The measured conductance and capacitance have been assumed to only be determined by
the layer itself, not by the interfaces and contacts,
Unknown growth rate of the ALD deposited alumina inside the quantum dot layer and on
top.
68
Chapter 6. Stack results
Further characterization of the latter is necessary to be able to correctly model all different
layers.
Based on the thick NC layer (70 nm), a relative permittivity of 4.52 is obtained. To compare this
experimental permittivity with the real value, a crude approximation can be done. Assuming
close-packing of equal spheres, built from a 7 nm radius CdS (εr,bulk = 5.7, ignoring the CdSe
seed of 2.2 nm) and a 2 nm shell of oleic acid (εr = 2.03), and the remaining 24 % filled with
alumina (εr = 7.5), an average relative permittivity of 4.7 is found [94, 95]. Both values are in
quite good agreement, showing that the model can certainly be used if an extensive study is
performed on the growth characteristics of the ALD alumina layer.
6.5 Electroluminescence of the AC stacks
The alumina layers of the stacks which should work under field-driven ionization can, as already
mentioned before, be deposited by electron beam evaporation or by ALD. The quality difference
can be confirmed by following observations during EL measurements:
Increasing the voltage precisely until light is emitted which is just visible by eye, induces
the defects of the e-beam layers to disappear very fast (blue sparks). However, their huge
amount evaporates the aluminium contact on top, resulting in failure of the device. Mostly,
this happens within the minute after the heredescribed process.
When repeating the same process for the ALD samples, only a very low number of defects
is removed when the voltage is turned on to the emitting point. The degradation of the
device occurs much slower, unless the voltage is significantly increased.
It should again be stated that light has been detected in both systems, but instability of the
e-beam samples prevents spectrometer measurements due to difficult alignment.
A first EL and PL spectrum of the stack ITO-Al2O3 (38 nm)-QDCdSe/CdS (1x SC, 35 nm)-Al2O3
(50 nm)-Al contacts is shown in fig. 6.20. The dots have a diameter of 14 nm and a QY of 22 %.
It should again be noted that the upper layer of 50 nm is in fact significantly smaller, since the
material penetrates and fills the empty spaces in the quantum dot/rod layer first before growing
uniformly on top of the dots/rods.
The EL spectrum has been plotted by subtracting the off spectrum from the on spectrum to
remove the unwanted background influence. Both spectra are clearly well aligned, signifying that
for the electroluminescence, the same radiative emission is present as for the photoluminescence.
The small blueshift of the EL spectrum might be caused by a slight calibration deviation of the
CCD camera.
The EL spectrum has been measured by applying a 20 kHz AC block voltage pulse of about
100 V peak-to-peak. The theoretical waveform is shown in fig. 6.21, together with the real
69
Chapter 6. Stack results
1.0
0.8
0.6
0.4
0.2
0.0
Nor
mal
ized
inte
nsity
[a.u
.]
700650600550500λ [nm]
PL EL
Figure 6.20: Electroluminescence spectrum of the stack ITO-Al2O3 (38 nm)-QRCdSe/CdS (1x SC)-Al2O3
(50 nm)-Al contacts.
voltage-time diagram.
-100
-50
0
50
100
volta
ge [V
]
6050403020100time [µs]
(a) Ideal block pulse. (b) Real block pulse.
Figure 6.21: Applied waveforms to measure electroluminescence.
As we obviously see, the steep rise in voltage has been strongly modified due to the capacitance
of the device load. The speed from low to high voltage is entirely determined by the slew rate
(expressed in V/µs). Unfortunately, the capacitor at the output load has to be charged to the
desired voltage as well. Therefore, if there is a limiting current IL of the amplifier itself, the
load capacitor can only be charged with a maximum speed. Thus, the slew rate is in this case
given by:
SA =ILCload
(6.10)
This value can be significantly lower than the unloaded one, especially in this case where capac-
itances of the order of tens of nanofarad are used [96].
Returning to the electroluminescence measurements, we can plot the intensity of the light in
function of time. This is done by measuring the surface under the spectrum peak at a certain
delay time after the trigger reaches the spectrometer. The result is shown in fig. 6.22. The ideal
block pulse is also shown to make interpretation easier. Two different curves are visible, which
are up to a factor almost equal to each other. The second (blue) curve is measured a few minutes
70
Chapter 6. Stack results
inte
nsity
[a.u
.]
50403020100t [µs]
-100
0
100
voltage [V]
Run 1 Run 2 Voltage source
Figure 6.22: Light intensity in function of time.
after the first one, leaving the circuit untouched. This indicates slight degradation of the device
during longer operation. The position of the peaks can be interpreted based on the theoretical
approach in section 3.2. After the steep rise of the voltage, the electric field induces moving
electrons and holes in the conduction and valence band respectively, resulting in an increase in
the light emission when they recombine radiatively. During the peak, the internal field builds
up and opposes the external one, resulting in evanescent emission. When the pulse is reversed
to the negative, the same mechanism occurs in the opposite direction. The peak at exactly one
period might be caused by the internal field itself, but no confirmation of this has been found.
Now that the power output of the dots is confirmed to originate from the electroluminescence,
power measurements are performed at different voltages and a fixed frequency of 20 kHz. For
the quantum CdSe/CdS dots, this is depicted in fig. 6.23.
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Pow
er [µ
W]
959085807570Voltage (pk-pk) [V]
Data points Linear fit
Figure 6.23: Output power of the AC stacks with CdSe/CdS dots (70 nm QD layer) in function of the
applied voltage.
The power output increases linearly with the applied peak-to-peak voltage. As already men-
tioned in subsection 6.4.1, the intercept of the linear fit crosses the horizontal axis at 72 V. The
71
Chapter 6. Stack results
slope of the fit yields 0.1 µW/V. The maximal measured power was 2 µW. Converting these
data to a power density yields:
Slope: 2.86 µW/Vcm2
Maximum power density: 57 µW/cm2
The latter value is significantly higher compared with values from literature one year ago [61].
Similar measurements have been performed on giant yellow-emitting CdSe/ZnS dots. The power-
voltage plot is shown in fig. 6.24. The same linear trend (with a different slope) and maximum
power output of 2 µW have been measured.
1.0
0.8
0.6
0.4
0.2
0.0
Pow
er [µ
W]
858075706560Voltage (pk-pk) [V]
Data points Linear fit
Figure 6.24: Output power of the AC stacks with giant CdSe/ZnS dots in function of the applied
voltage.
Power has also been measured for other types of dots/rods. For a layer of quantum rods emit-
ting yellow, a maximum power has been measured of 0.5 µW and breakdown occurred fast.
Measurements on the infrared PbS/CdS dots yielded a maximum power output of 0.3 µW.
Next, the power in function of the applied frequency (and constant voltage of 85 V peak-to-peak)
is plotted in fig. 6.25 for the case of the giant CdSe/ZnS dots.
A linear fit through the origin is performed on the first four data points, showing its propor-
tional increase. The power of the 20 kHz data point is lower than expected, because, as already
discussed, the shape of the block pulse is heavily deformed at this frequency. The deviation can
also partially be appointed to the internal electric field, which, at high frequencies, does not
have time to reach its full strength. This will be confirmed in the next section.
Based on the maximum power output of 2 µW at a voltage of 92 V peak-to-peak and a fre-
quency of 20 kHz, combined with the conductance G of the total stack, the efficiency η can be
determined, yielding:
η =Pout
Pin=
2 µW
GV 2rms
= 0.02 % (6.11)
Rewriting this value in lm/W by using the standard conversion 680 lm = 1 W (for a wave-
length of 555 nm) gives a power efficiency of 0.14 lm/W. This is only slightly lower than the
72
Chapter 6. Stack results
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
Pow
er [µ
W]
20151050f [kHz]
Data points Linear fit
Figure 6.25: Output power of the AC stacks with giant CdSe/ZnS dots in function of the applied
frequency.
values between 0.17 and 3.8 lm/W, obtained for DC LED stacks in literatrure [58,92]. Further-
more, the power output scales linearly with frequency and voltage, while the power input scales
quadratically with both. This results in the dependence:
η ∼ 1
ωV(6.12)
which should definitely be taken into account when considering steps towards commercialization.
To show the emitted light of the pads visually, pictures of the on and off state of one pad are
shown in fig. 6.26. The used nanoparticles are yellow-orange CdSe/CdS rods.
(a) On state. (b) Off state.
Figure 6.26: On and off state of yellow-orange emitting CdSe/CdS rods.
Two things can be concluded from the picture. On the one hand, only a major part of the surface
emits light, while the rods on the other part are degraded, probably caused by the removal of
the ligands during ALD deposition. On the other hand, white dots can be observed, which show
small breakdown paths through the device caused by the high voltage.
73
Chapter 6. Stack results
6.6 Modeling of the EL of the AC stacks
Based on the operation mechanism and the result of the intensity measurements during one
period, a Matlab model has been developed to confirm them. Therefore, a quantum dot layer
of three monolayers has been used, where only recombination of the central layer is considered
as the main contribution to the power output. Rate equations have been developed to match
the experimental behavior. For the electrons in the conduction band of the central layer, they
are modeled for the positive block pulse as:
nie,c =
ni−1e,c +
[ T1︷ ︸︸ ︷a (VQD − Eg/q) +
T2︷ ︸︸ ︷b(ni−1e,l − n
i−1e,c
)−
R︷ ︸︸ ︷cni−1e,c n
i−1h,c
]dt VQD > Eg/q
ni−1e,c +[b(ni−1e,l − n
i−1e,c
)− cni−1e,c n
i−1h,c
]dt 0 < VQD ≤ Eg/q
(6.13)
where a, b and c are constants which model the tunnel probability between the quantum dots and
the recombination rate. ni−1e,l and ni−1h,c respectively denote the electrons in the conduction band
of the left layer and the holes in the valence band of the central layer at time step (i− 1). The
term T1 models the overlap between the valence band of the left monolayer and the conduction
band of the central monolayer, which scales with (VQD − Eg/q). The factor a is dependent
on the drift mobility and tunnel probability between the two density of states. The term T2
denotes the transport between the conduction bands from the left to the central and the central
to the right layer, caused by drift, where b is naturally dependent on the drift tunneling mobility.
The final term R is the recombination term and is of course dependent on the electrons in the
conduction band and the holes in the valence band of the central layer. When VQD < Eg/q, the
overlap between the valence band of the left layer and the conduction band of the central layer
is zero. Of course, rate equations for the outer layers and for the holes in the valence band have
been developed with similar terms, both for the negative as the positive block pulse.
To show the agreement between the model and the experiment, the number of recombining
electron-hole pairs in the central layer is depicted in fig. 6.27 for three different frequencies.
The similarities of the peak and the evanescent intensity are clearly visible. Furthermore, they
confirm the linear increase of the power output in function of the frequency. Indeed, for the
frequencies f and 2f , the integral under the period of the output with frequency f is almost
exactly doubled when the frequency is doubled. However, deviations from this principle occur
when even higher frequencies are used, because the internal field does not yet reach its full
strength when the polarity of the external electric field is reversed.
Optimizing power output, the model might be expanded by considering other waveforms than a
block pulse, such as a sine or a triangle. Furthermore the time between subsequent pulses might
be changed, inducing lower frequencies, but more electric field alterations. Finally, varying the
number of layers might yield an optimal value for the power output.
74
Chapter 6. Stack results
0
x 1011
t [a.u.]
Inte
nsity
[a.u
.]
f2f4f
Figure 6.27: Number of recombining electron-hole pairs in the central layer in function of time and
different frequencies.
6.7 Comparison between both methods and commercialization
considerations
To conclude this chapter of the stack results, a summary of both operation mechanisms, their
advantages and disadvantages is given in table 6.1 and the possible commercialization is discussed
hereunder.
Quantum dots for light-emitting purposes have already been applied commercially by QD Vision.
Indeed, they have used them to enhance the emission from an LED replacement lamp and are
currently researching the possibility of commercial QD LEDs by direct charge injection [97].
Further, Samsung is investigating large-area full-color QD displays on glass and even flexible
substrates based on the same method [98]. The commercial use of the AC stacks, which require
high voltages and frequencies, is still in a research phase because of its fast degradation and
higher load on the electronics. However, polychromatic devices with phosphors operating under
voltages higher than 200 V and frequencies between 0.1 and 1 kHz, have already been designed
and commercialized by Planar [99]. Further successful research have been performed by iFire,
who developed full-color prototypes with phosphors, but only at voltages up to 60 V and a
120 Hz frequency [100, 101]. Luminescence experimentation for applications in displays has
been conducted in sulfides, ranging from rare earth materials to doped and undoped nanocrystals
[102].
75
Chapter 6. Stack results
Direct charge injection - DC Field-driven ionization - AC
Working principle
Advantages
Low DC voltages needed
Wide tunability in resistivity and mo-
bility of the layers
PL conservation for investigated lay-
ers
Big variety of layer structures to im-
prove the carrier injection, e.g. block-
ing layers, inverted device structure
Easy processing of the oxide layers
Easy characterization of the individ-
ual layers
Power output linearly tunable in
function of the applied voltage
Optimization possible by varying the
applied voltage waveform and the
number of layers, based on model
Disadvantages
Difficult optimization process due to:
– Wide tunability of a big set of
parameters
– Hard measurements e.g. band
alignment, carrier density, . . .
Lack of processing equipment to make
efficient organic injection and trans-
port layers
High AC voltage source needed at
kHz frequencies
Small tunability of the device itself
Reduced operation time due to high
voltages and frequencies
PL quenching for core-only quantum
dots
Table 6.1: Summary of both operation mechanisms, their advantages and disadvantages.
76
Chapter 7
Integrated light source under AC
field excitation
Integration of light-emitting devices in plasmonic structures or on silicon is a promising field.
The applications of integrated plasmonics are already discussed in the introduction, while the ad-
vantages of CMOS compatible silicon with the advanced functionalities of photonic components
are already widely tested. Indeed, applications of silicon in telecommunications, biosensing, etc.
are already known for their passive functionalities (signal routing, filtering, . . . ), yet the active
photonics such as on-chip light generation or detection proves to be more challenging, mainly
due to the indirect band gap of silicon.
This chapter begins with an introduction of the surface plasmon concept, followed by simula-
tions of the quantum dot, modeled as a dipole light source emitting in a metal-dielectric-metal
structure, which will be discussed and compared with literature. The focus mainly lies on the
deduction of the wavelength-dependent spontaneous emission enhancement factor. Further, the
practical integration of our bulk device on a silicon waveguide is described and experimental loss
measurements of different structures are interpreted.
7.1 Plasmonic integration
7.1.1 Surface plasmon concept
Plasmon is a term used to denote a quantized bulk plasma oscillation of electrons in a metallic
solid. When the limitation to surface plasmons is made, it refers to the interaction of the plasma
oscillations at surfaces. Therefore, its definition can be reformulated in a classical electromag-
netic model as a fundamental electromagnetic mode of an interface between a material with a
negative permittivity and a material with a positive permittivity having a well-defined frequency
and which involves electronic surface-charge interaction [103].
To arrive at a description of its propagation mechanism, the idealized situation of two materi-
77
Chapter 7. Integrated light source under AC field excitation
als with a real permittivity can be considered, where we start from the equations describing a
slab waveguide and letting the thickness of the central layer approach to zero. This yields only
appropriate solutions for the TM polarization with a propagation constant β given by:
β =ω
c
√εdεmεd + εm
(7.1)
This should be real, which is only possible, for the non-radiative surface mode, when εm < 0
and |εm| > εd. Inserting the expression for β in the appropriate equations gives us the magnetic
field: hy = A exp(−δx) x ≥ 0
hy = A exp(−γx) x ≤ 0(7.2)
where γ and δ are real, positive and function of the frequency and the permittivities of the
materials, while y is the direction parallel to the interface and perpendicular to the propagation
direction and x the direction parallel to the normal of the interface. The analysis can easily be
extended for a lossy metal. The shapes of the fields hy, ex and ez remain the same, characterized
by:
Exponentially decaying with a peak at the interface,
Much deeper penetration into the dielectric than into the metal.
The imaginary part of the permittivity of the lossy metal describes the loss of the surface plasmon
mode while propagating. If the real part is described by the Drude model, a non-radiating surface
plasmon can only exist when
ω <ωp√
1 + εd,real(7.3)
where ωp denotes the plasma frequency [104].
Now that the fields have been determined for a single metal-dielectric interface, extension can
be made towards an MDM structure. Indeed, if two interfaces are brought close together, the
resulting mode is the fundamental transverse magnetic mode, also called the gap surface plasmon
polariton mode, which is strongly confined between the two interfaces. Here, three parameters
can be considered as crucial in the surface plasmon propagation: the propagation length1, the
spatial extent2 and the confinement factor3. Therefore, all three are plotted in function of the
central layer thickness for gold-air interfaces and both the metal-dielectric-metal as the dielectric-
metal-dielectric stack in fig. 7.1. The free-space excitation wavelength is 1550 nm.
As expected, the spatial extent of the mode in the MDM structure is strongly reduced when
smaller than a critical value (12.5 µm), as is the case for the confinement factor. It is important
1Distance wherein the electric field intensity of a traveling wave at either surface decays by a factor of 1/e.2Distance between the points in the two cladding regions where the magnetic field decays to 1/e of its peak
value.3The ratio of power in the center region of the waveguide to the total power in the waveguide.
78
Chapter 7. Integrated light source under AC field excitation