Optical spectroscopic studies on semiconductor nanostructures SynopsisofThesissubmittedtothe TataInstituteofFundamentalResearch forthedegreeof DoctorofPhilosophyinPhysics AmlanMukherjee DepartmentofCondensedMatterPhysicsandMaterialsScience TataInstituteofFundamentalResearch Mumbai,India October2014 (FinalThesissubmittedinFebruary2015) Synopsis Semiconductor nanostructures continue to be of interest both for the fascinating array of novel physical phenomenon observed in them, as well as their potential for application in newer and improved electronic devices. This Thesis describes a set of optical spectroscopic studies carried out on semiconductor nanostructure systems in order to understand aspects of their electronic band structure (EBS) and its possible consequences for practical applications. We have studied group III-V as well as group II-VI semiconductors, where the system has nanometer size in 1- Dimension: such as Quantum Wells (QW), in 2-Dimension: such as nanowires (NW) and also in 3-Dimension: such as Quantum Dots (QD). In general, nano-sized semiconductors can exhibit quantum confinement related changes in the EBS and also other phenomena that affect the elec- tronicenergyspectrumsuchasenhancedexcitoneffects. Thatapartnanosizeformcanalsoresult in the formation of new crystal structures which are not found in bulk crystals, leading to a sig- nificant modification of the EBS. Understanding the EBS is crucial for the use of these materials for practical device applications. Optical spectroscopy is a powerful tool for investigating the EBS of such systems. In our work we have used a variety of spectroscopic techniques to study such nanostructured systems. Often, multiple spectroscopic techniques were employed to obtain complementary information. The work also involved building spectroscopic measurement setups and novel modifications to existing techniques to address specific problems being studied. The systemsstudiedhereexhibitcomplexopticalspectraandanalysisofthesemeasuredspectraoften involved first-principles lineshape simulations. The specific problems that were studied as a part ofthisThesisarelistedbelow: (i) Finding the origins of additional spectral features in the photo-reflectance (PR) spectra of GaAs/AlGaAs semiconductor QWs. This also involved building a novel detection scheme whereby the signal to noise ratio in a PR measurement could be improved by more than an orderofmagnitude. (ii) Determining EBS parameters of Wurtzite GaAs through polarized spectroscopic studies on single NWs. This also involved building an optical spectroscopy setup capable of measurements withhighspatialresolutionatlowtemperatures. (iii) Establishing the occurrence of a threshold in pump photon energy E beyond which photo- th luminescence (PL) throughput from CdSe-ZnS core-shell QDs do not increase although the ab- sorptioncoefficientcontinuestoincreaseandunderstandingtheoriginofthisphenomenon. iv (iv) Finding an explanation for the observed large absorption coefficient in CdSe-ZnS core-shell QDsathighphotonenergieshighabovetheireffectivebandgap. Theresultsfromthesestudiesareorganizedinfivechapters. Abriefdescriptionofthecontentsof thesechaptersisasfollows. Chapter1: Theoreticalconcepts Thischapterprovidesanoverviewofhowopticalpropertiesofnanostructured semiconductors are related to the EBS. Basic optical processes like absorption, reflection and emission depend on theaccessibleelectronicenergystatesinasolid. Thissubjectwillbetreatedunderthesemiclassi- calapproximationforlight-matterinteractions. Thestartingpointisadescriptionoftheabsorption process through the light field induced interband transitions between the valance band (VB) and theconductionband(CB)statesinthesemiconductormaterial. Fromtheabsorptionspectrumone canobtaintheimaginarypartk ofthedielectricfunctionandthenusingKramers-Kronigrelations the real part n. From n,k one can determine the reflection spectrum while the emission spectrum can be related to the absorption spectrum after taking into account the occupation probabilities. To understand reflection from multi-layered nanostructures one has to consider interference ef- fectsandthetransfermatrixapproachusedtocalculatetheopticalresponseofsuchmulti-layered structureswillbedescribed. Thiswillbefollowedbyadiscussionontheelectronicenergyspectrum. GroupIII-VandII-VI compoundsemiconductorsarefoundmostlyintwocrystalformsnamelyZinc-Blende(ZB)having acubicunitcellandWurtzite(WZ)whichhasahexagonalunitcell,asshowninFig.1. Thecrystal structure affects both the relative energy positions of the bands as well as the composition of the cell-periodicpartsofthewavefunctionsassociatedwithanenergyband. Inthesesemiconductors, at the Brillouin zone center (crystal momentum k=0), the cell periodic part of the conduction band (CB) wavefunction has atomic s orbital like character and those for valence bands (VB) have a mixture of atomic p , p or p orbital character as indicated in Fig. 1. A dipole transition x y z betweensandp stateswillinvolvex-polarizedlightandsoon,whichgivesrisetothepossibility x of linear polarization sensitivity of the transitions depending on the p , p , p contribution to x y z the VB [1]. In bulk cubic crystals the high symmetry of the lattice is reflected in their EBS, making their optical properties insensitive to the polarization of light. Semiconductors with WZ crystal structure have an unique c-axis defined normal to hexagonal base of it’s unit cell. This lower symmetry suggests the possibility of polarization sensitive optical transitions relative to the v z [001] z [0001] c-axis Zinc Blende Wurtzite y [010] z x [100] y [1100] y x [1120] x CB s CB z zz zz HH A y y y ergy LH x x x ergy B n n E k k=0 SO px py pz E k k=0 C FIG.1. UnitcellofZinc-Blende(ZB)andWurtzite(WZ)crystalstructureandthenatureoftheelectronic band structure around the center of the Brillouin zone of unstrained group III-V semiconductors having such crystal structures. The atomic orbitals that contribute to formation of the conduction band (CB) and thethreevalancebandsZB(heavy-holeHH,light-holeLHandspin-orbitsplitoffholeSO)andWZ(A,B andChole)areindicated. c-axis. The crystal structure of our nanostructured semiconductors will play an important role in determining their EBS. We will briefly discuss the theoretical formalism used to understand how crystal structure influences the EBS resulting in linear polarization anisotropy in the optical response. When the size of a semiconductor is reduced to nanometer length scale so that it is compa- rable to the de-Broglie wavelength of charge carriers having kinetic energy (cid:39) k T, effects of B quantum confinement become prominent. Such systems include quantum wells, quantum wires and quantum dots. We will describe how quantum confinement affects the EBS which results in an increase of the effective bandgap and modification of the joint density of states. For instance quantum confinement in 3-dimensions give rise to the possibility of discreet atomic spectra like transitions. Nanostructuring also affects carrier-carrier interactions. In particular we would be interested in excitons which can be thought of as “hydrogen atom like” bound state of an electron and hole pair due to Coulomb interaction. Excitonic effects give rise to spectral features below the bandgap in an absorption spectrum and also modify the absorption spectra above the bandgap [2]. The exciton energy spectra can be modified in nanostructured semiconductors, for instance vi the exciton binging energy increases significantly in a QW. Changes in the EBS due to quantum confinement and due to modification of exciton properties in semiconductor nanostructures will bedescribedinthischapter. Chapter2: Experimentaltechniques This chapter describes the experimental setups built and the various spectroscopic techniques used in our studies. These include reflectance, transmittance, conventional photoluminescence (PL)aswellasspatiallyresolvedmicro-photoluminescence(µ-PL)andmicro-Raman,photolumi- nescenceexcitation(PLE),andphotoreflectance(PR)spectroscopy. Forthemeasurementsthatre- quiredamonochromaticprobebeam,reflectiongratingbasedMonochromators(MC)wereusedto dispersewhitelightfromXenondischargelampsorconventionalTungstenfilamentlamps. Lasers withphotonenergyhigherthanthebandgapofthesampleunderstudywereusedforexcitationin PLandPRmeasurements. Whilesiliconphoto-diodeswereuseddetectsignalofnormalintensity, thermo-electrically cooled silicon charge coupled device (CCD) was used to detect low intensity emission signals during PL and PLE measurements. PR measurements requiring phase sensitive detection were carried out using Lock-in amplifiers. A Glan-Taylor, Glan-Thomson or polymer- film polariser was used for polarization selective measurements. Low temperature measurements were accomplished using either a continuous-flow liquid He cryostat (4.5 K) or a closed-cycle He cryostat (10 K). Programs were written in LabVIEW software for data acquisition and instru- mentcontrolforallthemeasurementsetups. Thesemiconductorcrystalsusedinourstudieswere obtainedfromourcollaborators. Webrieflydescribedbelowthemainspectroscopysetupsused. PL spectroscopy provides useful information about the EBS around the fundamental bandgap which includes the bandgap energy, defect states as well as nature of electronic wavefunctions associated with the states involved in the transition. µ-PL spectroscopy combines microscopy with PL spectroscopy to achieve high spatial resolution. In a conventional PL measurement a relatively large area is probed which might contain a large number of nanostructures, for exam- ple NWs, under study. This results in an ensemble averaging of the spectra which can smear out distinct spectral features associated with individual NWs. µ-PL spectroscopy with its high spa- tial resolution is therefore useful for studying individual NWs so that the spectrum is free from ensemble averaging effects. The microscope part of the setup was built around a high magnifica- tion (100x, NA=0.5) long-working-distance (12 mm) objective. It was coupled to a 0.55 m focal lengthMCandthesignalwasdetectedusingathermo-electricallycooledSi-CCDwith512(rows) vii Lamp Camera Shutter LL22 L3 LPF Polarizer L4 Monochromator BS1 (movable) L1 NDF BBSS22 Shutter Laser CCD BPF BS3 (movable) LWDO Temperature Controller Sample Liq. He dewar in cryostat Vacuum pump station FIG.2. Schematicdiagramofthemicro-Photoluminescence/Ramanspectroscopysetupbuiltasapartofthis Thesis work. The abbreviations stand for following L: lens, BS: beam splitter, LPF: low pass filter, BPF: band pass filter, NDF: neutral density filter, LWDO: long-working-distance objective and CCD: charge coupleddevice. x 1024(columns) pixels. The setup also had a separate camera to see and align the sample pre- cisely. A schematic of the setup is shown in Fig. 2. During measurements, a NW was imaged on the entrance slit of the MC. The gold coated Si substrate on which the NWs were dispersed was carefully oriented so that image of the NW on the MC slit was aligned with its axis parallel to the entrance slit and an entire NW could be imaged on the CCD. To see such images of the NW we set the MC grating to zero-order position which results in a image of the entrance slit forming on the CCD with the grating acting like a simple plane mirror. By using an extra lens in the path of the laser (He-Ne 633 nm), the excitation laser spot on the sample was defocused so that the entire NW was uniformly excited. With this arrangement, the readings from a row of pixels on theCCDcorrespondstoemissionspectrumofaspecificsectionalongthelengthoftheNW.Thus readingoutandstoringsignalsfromeachrowofpixelsseparately,enabledrecordingtheemission spectrum of different sections along the length of the NW at one go, with a spatial resolution of ≈ 1.5 µm. The polarization of the emitted PL signal was resolved into components parallel ((cid:107)) andperpendicular(⊥)totheNWaxisusingaGlan-ThomsonpolariserplacedinfrontoftheMC. Using a sharp cut-off long pass interference filter for the laser, we could also perform µ-Raman spectroscopywiththissetup. WealsodesignedaPLEmeasurementsetupforstudyingQDswhere viii we made use of polarized excitation and detection in order to efficiently suppress the scattered pump beam from reaching the detector and thereby eliminated spurious signals. This enabled us toestimatetheabsorptionspectrumofQDsveryclosetothePLEdetectionwavelength. PR spectroscopy is an important tool for probing the EBS, especially at energies above the bandgap,and hasbeen extensivelyusedfor studyingsemiconductorQWs. In PRone perturbsthe sampleusingapumpbeamwhichistypicallyalaserthatgetsabsorbedandgenerateselectron-hole pairs which in turn modify the built-in electric fields at the surface or interfaces. The consequent dielectricfunctionchangeresultsinarelativereflectivitychange∆R/R,whichismeasuredusing a second probe beam of variable photon energy [3]. Typically ∆R/R ∼ 10−2 −10−4 and peaks around energies where critical points occur in the joint density of states, helping determine their energy. In a PR measurement the weak signal usually rides on a large background signal arising from PL, especially at low temperatures. Drift and fluctuations in the laser intensity causes varia- tionsinthisbackgroundwhichcanbelargecomparedtothePRsignalandcanoverwhelmitifthe PRsignalisweak. WedevisedanovelmethodfornullifyingthePLbackgroundsignalinrealtime [4]. This involved a detection scheme with two detectors, one measured the PL + PR signal while the other measured just the PL signal. The signal from the two detectors were subtracted in real time after providing electronic compensation to account for the difference in the amplitude and phase response of the two detection channels. In the resultant PR measurement the Signal/Noise ratio improved by more than an order of magnitude. The above techniques and others [5] used in ourworkwillbedescribedinthischapter. Chapter3: Systemswithnanostructurealong1-Dimension: GaAs/Al Ga AsQuantumWells x 1−x This chapter describes a study on single QWs which have nanometer size in 1-Dimension in terms of the QW width. The EBS changes significantly when carrier motion is constrained along 1-Dimension in a QW, giving rise to confined electron and hole states. In a GaAs/Al Ga As 0.3 0.7 QW, PR spectral features are expected at the onset of transitions involving confined electron and heavy-hole (e hh ) or light-hole (e lh ) states, where m, m’ are the confinement indices. In a m m(cid:48) m m(cid:48) QW carriers are free to move only in the 2-Dimensional QW plane which also affects the exciton energy spectrum. Due to increased exciton binding energy E , transitions in a QW typically in- b volveexcitons. TheincreaseE givesrisetothepossibilityofobservingfeaturesnotjustduetothe b groundstateexciton(n=1,nbeingtheexcitonlevelindex)associatedwitheache hh ore lh m m(cid:48) m m(cid:48) transition, but also due to higher excitonic states (n=2 and so on) [6]. In addition there have been ix Wavelength (nm) 760 755 750 745 740 4 e hh e lh 12 K 1 1 1 1 Fit x 5 2 3 0 1 x 0 R R/ -2 Δ -4 -6 1.62 1.63 1.64 1.65 1.66 1.67 1.68 Energy (eV) FIG. 3. Photoreflectance spectrum of a 4 nm thick GaAs/Al Ga As QW at energies around the first 0.35 0.65 FIG.1: Amlan Mukherjee et al. confinedelectron-heavyhole(e hh )andelectron-lighthole(e lh )excitontransitions. Additionalweaker 1 1 1 1 featuresseenonthehighenergysideofthemaintransitionsareidentifiedbyarrows. suggestionsaboutthepossibilityofaspectralfeatureattheexcitoncontinuumedge[7],wherethe n=∞excitonictransitionoccursandwhichcoincidesinenergywiththeonsetofinterbandtransi- tions. Thatapart,therecanbetransitionsthatarenormallyforbiddenbysymmetryconsiderations butariseiftheQWsymmetryisdisturbedbyperturbationssuchasbuilt-inelectricfields[8]. This study[9]dealswiththeobservationofadditionalweakPRfeaturesatenergiesslightlyhigherthan the e hh and e lh ground state (n=1) exciton energies in a GaAs/Al Ga As QW and finding 1 1 1 1 0.3 0.7 theirorigins. SuchaspectrumisshowninFig.3. Findingtheoriginsofsuchfeatureshasbecome important in the context of understanding the optical response of the new class of 2-Dimensional semiconductorssuchasMoS . 2 The GaAs/Al Ga As QWs used in the study had well width of 4 nm and were grown using 0.3 0.7 metal-organic vapor-phase epitaxy in the laboratory of Prof. S. K. Shastry. The PR experiments involved a special dual detection technique [4] described previously which enabled measurement of lineshapes of these weak additional features precisely. PLE spectroscopy measurement was also done on the sample, which showed broadened step-like features around these energies. A detailedlineshapeanalysis,includingfirstprinciplessimulations,wasperformedtounderstandthe originsoftheseadditionalPRspectralfeatures. Itincludeddeterminingtheabsorptionspectrumof inhomogenously broadened exciton transitions in 2-Dimensions and then using Kramers-Kronig transformations to ultimately obtain the real and imaginary parts of the refractive indices. There- x after the quantum-confined Stark effect was invoked to simulate the PR spectrum [10]. The analysis showed that these additional features arise primarily from inhomogeneously broadened first excited state transition of the excitons, rather than from a change in the joint density of states attheexcitoncontinuumedge. Theanalysisalsosuggeststhatsuchfeaturesaremorelikelyinthe case of 2-Dimensional excitons as compared to 3-Dimensional excitons in bulk material. Apart fromitssignificanceforpost-growthcharacterizationofQWs,withtheidentificationofthesetwo additional PR features with excited state exciton transitions with n=2, one can directly estimate the effective Rydberg constants and E which was found to be in good agreement with the ones b estimated from fitting the PLE spectrum. On the issue of whether these features have any direct connection with the exciton continuum edge, our analysis negates such a possibility. This is be- cause the joint density of states in 2D is continuous at the exciton continuum edge, as it is in 3D, and therefore a feature in the modulated reflectance spectrum related to the exciton continuum edgecannotarise. Chapter4: Systemswithnanostructurealong2-Dimensions: GaAsNanowires ThischapterdescribesastudyonsingleGaAsNWswhichhavenanometersizein2-Dimensions in terms of the NW diameter. As such semiconductor NWs are of interest due to their potential for application in high-density electronics and opto-electronics. In nano-size form materials can exhibit polytypes which are not present in the bulk [11]. For example in the case of GaAs, NWs canfromwithWurtzite(WZ)crystalstructurewhichisnotfoundinbulk,sincebulkGaAsalways crystallizes in the Zinc-Blende (ZB) structure. The band structure of WZ GaAs is currently under debate [12–14]. The band alignment of the WZ and ZB sections in a mixed phase GaAs NW is predictedtobeoftypeII[15]. SuchabandalignmentinmixedphaseGaAsNWsoffersinteresting possibilitiesforthequantumconfinementalongtheNWaxis,forwhichonenormallywouldhave to grow a different material with a lower band gap in the middle of the NW. We investigated the EBSofsingleGaAsNWswhichhavebothWZandZBsections,usingpolarizationresolvedµ-PL spectroscopy[16]. The GaAs NWs used in the study were grown using molecular beam epitaxy [13] at the Paul Drude Institute (Berlin). They typically have a diameter of 150 nm with length of 8-10 µm and are covered by a thin shell of Al Ga As to passivate surface states and thereby increase PL 0.1 0.9 output. These NWs were broken off and dispersed on a Au coated Si substrate for spectroscopic measurements. Results of µ-PL spectroscopy study on such a NW is shown in Fig. 4. The xi Low High 0 130 (a) (b) (c) 2 z (i) 4 140 ) m # h (μ 6 xel engt 8 (i i) 150 Pi L 10 12 160 1.45 1.5 1.55 1.45 1.5 1.55 EEnneerrggyy ((eeVV)) FIG. 4. (a) Diffraction limited image of a GaAs NW under white light illumination seen through the monochromator (MC) input slit. The image was taken with the CCD attached to the MC with its grat- ingsettozero-orderposition. SpectrallydispersedPLimageoftheNWforemissionpolarization(b)E⊥z and (c) E(cid:107)z, where z is along the NW length. The pixels along the vertical direction represent position alongtheNWlength. Inplots(b)and(c)pixelsalonghorizontaldirectionrepresentemittedphotonenergy and have been transformed accordingly. The sample temperature and excitation intensity were 4.5 K and 1.8 kW/cm2, respectively. In (b) and (c), Log of the PL signal count (scale indicated by bar on top) has 10 beenplottedtosimultaneouslyshowupweaklyandstronglyemittingregions. Thedashedhorizontallines markhighlyluminescentsectionsalongtheNWlengthfromwhichtheemissionis(i)strongly(ii)weakly polarized. PL is from localized regions along the NW. At low excitation intensity (5.6 W/cm2) the PL is strongly polarized perpendicular (E⊥z) to the nanowire axis along z. At high excitation intensity (1.8 kW/cm2) one observes regions along the nanowire length from which the PL continues to be strongly polarized with E⊥z (dashed white line indicating region (i) in Fig. 4), and also regions fromwhichthePLisnotstronglypolarized(dashedwhitelineindicatingregion(ii)inFig.4). The strongly polarized PL from region (i) at high excitation has higher average photon energy, with the E(cid:107)z emission occurring at a higher energy than the emission with E⊥z. This can be seen in Fig. 5. We considered various factors that can cause such polarization anisotropy in the measured emission from such NWs. The measured transition energies and polarization characteristics were analysed by comparison with EBS calculations. The strongly polarized emission at high energies is identified as arising from predominantly WZ regions of the NW, while the mostly un-polarized emission comes from the predominantly ZB regions. In WZ GaAs the top two valance bands xii