PhysRevB[1].81.115304

Competition between exciton-phonon interaction and defects states in the3.31eV band in ZnOD.Tainoff,1,2,3B.Masenelli,1,2,3,*P.Mélinon,1,2,3A.Belsky,1,2,4G.Ledoux,1,2,4D.Amans,1,2,4C.Dujardin,1,2,4N.Fedorov,5,†and P.Martin51Universitéde Lyon,Lyon F-69003,France2UniversitéLyon1,Villeurbanne F-69622,France3LPMCN,CNRS,UMR5586,Villeurbanne F-69622,France4LPCML,CNRS,UMR5620,Villeurbanne F-69622,France5CELIA,UniversitéBordeaux1–CNRS–CEA,351cours de la Libération,F-33405Talence Cedex,France͑Received23November2009;revised manuscript received22January2010;published4March2010͒We investigate the origin of the band at3.31eV͑A band͒commonly observed in the emission spectra ofvarious ZnO samples.This band is of prime importance for the confirmation of p doping in ZnO nanostruc-tures.We check the validity of the three main hypotheses generally evoked to explain it in undoped ZnO,namely,surface states,the1LO phonon replica of the free exciton and a defect-related ing ZnOsamples structured at different scales,from macro to nano through meso͑i.e.,a single crystal,a nanoparticlesassembly,and a microcrystalline pellet͒,we demonstrate that a huge surface/volume ratio does not necessarilyimply a strong emission at3.31eV,especially for model nanoparticles which are uncapped and synthesized inultrahigh vacuum.Furthermore,we show that the two other hypotheses are valid and can be at stake concomi-tantly,according to the quality of the samples.Regarding the1LO phonon replica of the free exciton,itspresence is unambiguously established using a complete model based on an exciton population at thermody-namic equilibrium,including the treatment of the interaction of the excitons with the acoustic phonon bath.Weobserve that the1LO phonon replica becomes significant at temperatures higher than80K typically.Below thistemperature,the3.31eV emission is only present in the microcrystalline sample and results from a defect-related transition.Since it is not observed in the nanoparticles that are made from the microcrystals,thepossibility of an impurity to be the origin of the invoked defect is unlikely in our experiments.Instead,ourstudy suggests that the defect at stake is of crystalline origin,with an activation energy of122Ϯ5meV.Therelated emission is shown to follow a free-to-bound transition mechanism.DOI:10.1103/PhysRevB.81.115304PACS number͑s͒:78.67.Bf,78.55.Et,71.55.GsI.INTRODUCTIONIn the present decade ZnO has been widely studied in sight of optoelectronic applications.Indeed the high stability of its room-temperature UV excitonic emission,its nontox-icity and its low cost of production make ZnO a good can-didate for the realization of UV optoelectronic compounds. However the main barrier to overcome in order to industri-alize ZnO optoelectronic compounds is to engineer the dop-ing properties and especially the p-doping one.In this con-text,the possibility for some features of the low-temperature excitonic luminescence to be the optical signature of the p doping is a major issue.This is particularly true regarding the controversial assignment to p doping of the3.31eV band ͑often called A band͒present in the excitonic spectra of vari-ous ZnO samples.Since the energy of the3.31eV band is close to the one expected for the optical phonon replica of the free exciton ͑FX͒,this band has beenfirst interpreted in this way.On the other hand,the appearance of this band in various kinds of nanostructured samples as dots,1,2nanorods,3or nanowires4 has conducted some authors to interpret this feature as a surface contribution.Finally,theoretical studies of some ac-ceptors energies in ZnO have shown that the3.31eV band might be related to the presence of acceptor impurities ͑hence the name“A band”for acceptor band͒through differ-ent radiative mechanisms like a donor acceptor pair,a free to bound transition or an exciton bound to a defect.Subse-quently this band has been seen in a lot of intentionally p-doped ZnO samples and it has been tempting to attribute it to the optical signature of p doping.5–8However the possi-bility for the acceptor level emission to be due to uninten-tional impurities or local crystalline symmetry modifications complicates the rigorous assignment of this band to p dop-ing.Since the control and the optical signature of the p dop-ing is of prime importance,the nature of the3.31eV band is of prime importance too.In order to study the origin of the3.31eV band we have made a systematic study of the temperature dependence of the excitonic luminescence of three kinds of ZnO samples which are structured at different scales:from macro to nano through meso scale.The three samples used are a single crystal͑macroscale͒,a microcrystalline powder͑mesoscale͒, and an assembly of ZnO nanoparticles͑nanoscale͒which are uncapped,controlled in stoichiometry,and crystallinity and deposited in ultrahigh vacuum͑UHV͒.The nanoparticles be-ing made from the microcrystalline powder by a hyper-quenching process,we expect tofind embedded in them the same impurities as in the microcrystalline powder if they are present in the latter.This enables us to discriminate the in-fluence of impurities from pure size effects,and among the latter the role of the surface states.The emission of these samples is compared to the one of a ZnO single crystal, which is characterized by its long-range crystalline order. Our results show that a huge surface/volume ratio does not necessarily lead to an important emission at3.31eV since no such emission is visible in the nanoparticle spectra at10K.PHYSICAL REVIEW B81,115304͑2010͒Second,the accordance of the temperature dependence of the 3.31eV band in the single crystal and in the nanoparticle assembly and itsfit by a model describing the exciton popu-lation at thermodynamic equilibrium,including the influence of the acoustic phonons show that the1LO phonon replica of the free exciton͑1LO-FX͒is distinguishable in the3.31eV band of ZnO.Eventually,the comparison of the lumines-cence of the microcrystalline powder at10K with the spectra of the two other types of samples,for which the band is absent at the same temperature,reveals that,on top of the 1LO phonon replica,the3.31eV band also originates form a defect state͑real A band͒.This feature is ascribed to a free-to-bound͑FB͒transition.II.EXPERIMENTSThe three kind of samples studied are a single crystal͑SC͒from SurfaceNet Gmbh,a microcrystalline͑␮Cs͒pellet which has been obtained by sintering͑pressing at15ton and subsequent annealing at1000°C in O2atmosphere for10h͒the99.999%pure ZnO powder from Cerac and a nanopar-ticles assembled thinfilm synthesized by low-energy clusters beam deposition͑LECBD͒.9The nanoparticles͑NPs͒are synthesized by the supersonic quenching of a plasma ob-tained from the laser ablation of the microcrystal pellet.The nanoparticles are synthesized in the gas phase,with the ad-dition of O2in the carrier gas,and then deposited in UHV conditions.The nanoparticles study by transmission electron microscopy͑TEM͒and x-ray diffraction͑XRD͒shows that the NPs are crystallized in the wurtzite ing Scherrer’s formula,a mean size of18nm is deduced from the full width at half maximum of the XRD peaks corre-sponding to the͑100͒,͑002͒,and͑101͒plane families.Fig-ure1presents the XRD pattern of thefilm.Two TEM images are also given as an illustration of the individual particles and their assembly that occurs within thefilm formation, leading to diffracting domains about18nm wide as mea-sured with XRD.The stoichiometry of the NPs,controlled in situ using x-ray photoelectron and Auger electron spec-troscopies,is identical to the␮Cs one.In both cases no im-purity is identified in an amount of1%or more͑the precision of our XPS setup͒.In addition of the good crystallinity re-vealed by TEM and XRD,no green luminescence is ob-served in the NPs cathodoluminescence spectra.This green luminescence is assigned to surface contamination,10oxygen vacancies,11,12or Cu impurities.13More details on the NPs synthesis process,stoichiometry and crystallinity are pub-lished elsewhere.14The excitonic emission of the samples has been recorded with the experimental setup described in Ref.15.However,in the present case,we have used the300 nm͑4.13eV͒radiation of a Xe lamp as the excitation source, a1200groves/mm grating blazed at300nm and a CCD camera of TRIAX320spectrometer͑Jobin Yvon͒.The setup resolution we have used was1meV.Unless specified,all the spectra shown in this study are normalized to the emission maximum.III.RESULTSIn the following,we address successively the three main hypotheses evoked to explain the origin of the3.31eV band,which are,as mentioned in the introduction,the role of sur-face states,the1LO-FX replica and the role of bulk defects, either linked to impurities or to structural faults.A.Role of the surfaceThe enhancement of the3.31eV emission in various ZnO nanostructures,which are characterized by a large surface/ volume ratio,has lead to assign its origin to surface states.1,16 In this sense,this emission would result from a geometrical-size effect.In order to probe this hypothesis,we compare the excitonic luminescence of a ZnO single crystal and twofilms of ZnO nanoparticules,one protected by a200-nm-thick MgO layer͑deposited in UHV͒and one unprotected.Figure 2presents the spectra corresponding to two representative temperatures,namely,10and80K.At10K,in all spectra, the excitonic emission is strongly dominated by the donor-bound emission bands called DXs.17–19In the3.33eV range, the emission exhibits a weak two electron satellite,which involves a DX exciton leaving the donor in an excited state, and some contribution of defects,17especially in the case of the MgO-protected nanoparticlefilm.However,the investi-gation of this emission is beyond the scope of the present paper.All these peaks have been successfully assigned with respect to the literature.11,12In the3.31eV region,no emis-sion is detected for any sample.At80K,the FX contribution increases to the detriment of the DXs one,corresponding to the thermally activated detrapping of the donor-bound exci-ton.Concomitantly,the3.31eV band appears for the two kinds of samples.The absence of the band in the NPs spec-trum at10K and its moderate intensity at80K seems in contrast with other studies.11,18Actually,only a few studies are rigorously concerned with the increase in the3.31eV emission yield as the surface/volume ratio increases.1,2In most of them,the size of the nanostructures is typically a few tens or hundreds of nanometers,which leads in the best es-timate to ratios of about0.1%.In the case of our nanopar-ticles,this ratio is more than ten times larger.Therefore,we would expect them to exhibit a significant band at3.31eV, which is not the case.Two hypotheses can be evoked to explain this fact:either our peculiar synthesis method͑adia-batic supersonic expansion͒prevents the formation of the surface defects that may be responsible for the3.31eV band or it induces the formations of a majority of nonradiative surface defects with a decay time shorter than the one of the 3.31eV emission which results in the quenching of this emission.Indeed,it is well known that the outermost layer of a nanostructure͑for instance,a shell whose thickness is about the exciton diameter͒can be the place of many defects that can have opposite influences.In the present study,it is difficult to tell which hypothesis can prevail.In order to characterize the nanoparticle surface at best,we remind that, whereas the samples of Fallert et al.are made by“compress-ing industrial powders intoflat pellets”1and those of Fonoberov et al.2result from wet-chemistry routes,the present NPs are uncapped͑no ligand͒and their surface has been formed during a drastic cooling process͑ϳ1010K/s͒and possibly reconstructed in UHV.Their subsequent analy-sis at air is bound to have induced an adsorption of OHTAINOFF et al.PHYSICAL REVIEW B81,115304͑2010͒groups on their surface as is often the case with polar oxide nanoparticles.20–22We have checked that the OH contamina-tion does not change the excitonic emission.23Besides,as can be seen on Fig.2,protecting the clusters by an MgO layer does not modify the spectra in the 3.31eV range.As already mentioned,the only change concerns the defects states responsible for the emission at 3.33eV .To sum it up,our investigations go to show that increasing the surface/volume ratio does not imply a monotonic increase in the 3.31eV emission.The discrepancy with other studies seems to be only apparent since the presence,nature and role of surface defects depend on the synthesis method.That is the reason why we only claim that a naked ZnO surface,possibly re-constructed in UHV ,does not induce an emission at 3.31eV .B.1LO phonon replicaHighly polar semiconductors,such as ZnO,are known for their significant exciton-phonon coupling.24The coupling is strong enough to allow the observation of polaritons even at moderate temperature.25,26It is thus sound to contemplate the 1LO phonon replica of the free exciton as the potential origin5nm5nmFIG.1.Upper panel:XRD pattern of the nanoparticle film ͑ex-cited by the Cu K ␣radiation ͒.The wurtzite structure is identified by the ͑100͒,͑002͒,and ͑101͒peaks.The size of the diffracting do-mains is deduced from the full width at half maximum of the three peaks using Scherrer’s equation.Middle panel:TEM image of an individual particle,the unit building block of the nanostructured sample.Lower panel:example of the assembly of several nanopar-ticles.Coherent crystalline domains,over 15nm wide,are clearly distinguished.The scale bar represents 5nm in the two TEM images.Energy (eV)Energy (eV)FIG.2.͑Color online ͒Upper panel:emission spectra at 10K of ͑from bottom to top ͒a ZnO single crystal,an assembly of unpro-tected nanoparticles,an assembly of nanoparticles protected by a thin MgO film.The excitation wavelength is 300nm.The absence of the band at 3.31eV in the nanoparticle sample indicates that its origin is not intrinsically linked to the ZnO surface.Lower panel:emission spectra of the same samples at 80K.The 3.31eV band becomes observable.COMPETITION BETWEEN EXCITON-PHONON …PHYSICAL REVIEW B 81,115304͑2010͒of the 3.31eV band.The values of the 1LO phonon energy and of the FX energy,being,respectively,72meV and 3.37eV around 80K,give credit to this hypothesis.One last observation that supports the assignment of the 3.31eV band to the 1LO-FX replica is the evolution of the band shape with temperature.As shown on Fig.3for the single crystal,the 3.31eV band tends to be more asymmetric on its high-energy side as the temperature increases,which is commonly observed for phonon replicas.4,27,28For the particular mea-surement of the 3.31eV band evolution with temperature,in order to increase our resolution,we have used a continuous HeCd laser ͑325nm ͒of 0.1mW for excitation and the spec-tra have been recorded with an intensified CCD camera coupled to a 3600groves/mm grating ͑ARAMIS confocal setup from Jobin Yvon ͒.The resolution of this setup is 0.15meV and permits us to interpret our line shapes unambigu-ously.The results obtained from the SC sample are shown in Fig.3for three different temperatures.Beside the main DX and FX contributions merged at 3.36eV ,we observe the band around 3.31eV and another one around 3.23eV .These two bands can be assigned to the 1LO and 2LO phonon replicas of the free exciton as described hereafter.The most complete model accounting for the exciton-phonon interac-tion has been given by Segall and Mahan 28in the formalisms of the Green’s functions and even takes into account the polariton bottleneck.However,for an exciton gas in thermo-dynamic equilibrium,the approach due to Permogorov,29which considers a Maxwellian distribution for the kinetic energy of excitons leads to identical results in most of the cases.30According to this simplified model,the shapes of the 1LO-FX and 2LO-FX are given,respectively,by the follow-ing formulas,31where E 1LO is the energy of the 1LO phonon and E 2LO =2E 1LOE ͑ប␻͒ϰ͓ប␻−͑E FX −E 1LO ͔͒3/2exp ͕−͓ប␻−͑E FX−E 1LO ͔͒/kT ͖,͑1͒E ͑ប␻͒ϰ͓ប␻−͑E FX −E 2LO ͔͒1/2exp ͕−͓ប␻−͑E FX−E 2LO ͔͒/kT ͖.͑2͒The only free parameters are the intensities of the two peaks and the value of E 1LO .Figure 3shows the result of this modeling applied to the SC spectra for different tempera-tures.The shape of the peak obtained is in qualitative agree-ment for both the 3.31eV band and its 1LO replica for temperatures over 80K.However there is a cutoff in the shape which prevents the modeling of the low-energy part of the peaks.This limitation has been pointed out as soon as the introduction of the exhaustive model by Segall and Mahan.28Moreover the result from this modeling is not asymmetric enough to well fit the high-energy tail of both bands.As the discrepancy of asymmetry becomes more pronounced when the temperature increases,we have taken into account the contribution of acoustic phonons,which are a well-known source of broadening.The resolution of our setup being finer than the FX-band full width at half maximum,we can deter-mine the temperature dependence of the acoustic phonons contribution through the homogeneous broadening of the FX transition.In our case FX is fitted with a Lorentzian curve ͑cf.Fig.2͒and its energy distribution is convolved with the two previous equations in order to obtain a modified Per-mogorov law taking into account the homogeneous broaden-ing due to acoustic phonons.This procedure has been pro-posed several times 28,32but applied only recently.33The result of this procedure is presented in Fig.3.The modified Permogorov model is indisputably better.34It is also clear that this approach is quite general and can be used for other semiconductors.This modeling gives us the temperature de-pendence of the 1LO and 2LO phonon replica energies,which follows the FX temperature dependence.The E 1LO energy obtained at different temperatures for the optical phonons is within the 71–73meV range ͑cf.Fig.4͒which is in good accordance with both the literature values of 1LO energy determined by Raman spectroscopy 35,36and our Ra-man measurements ͑not presented ͒.This result proves that,if no band is detected at temperatures lower than 80K,the 3.31eV band and its 1LO replica that appear at temperatures higher than 80K are,respectively,the 1LO and 2LO phonon replica of the free exciton.Let us emphasize that all the previous analysis holds also for the NPs sample.C.Defect-state contributionFigure 5presents the emission of the microcrystalline sample at 10and 80K ͑before its use for ablation ͒.ContraryEnergy (eV)I n t e n s i t y (a .u .)FIG.3.͑Color online ͒Emission spectra of a ZnO single crystal ͑solid line ͒showing the 3.31eV band and its 1LO replica.The excitation is the 325nm radiation from a HeCd laser.The two features are modeled as the 1LO and 2LO phonon replicas of the FX contribution by a Permogorov law ͑cross curve ͒and a modified Permogorov law taking into account the broadening due to acoustic phonons ͑circles ͒.The modeling is presented for 80,100and 140K.The cutoff inherent to the Permogorov law is highlighted by a doted circle on the simulation curve for 140K.Note the improved accu-racy in the 3.30eV region when the Permogorov law is modified by taking into account the interaction of the free exciton with the acoustic phonon bath.PHYSICAL REVIEW B 81,115304͑2010͒to what is observed for the single crystal and the nanopar-ticles at 10K,the peak is observed around 3.31eV for the microcrystalline sample which cannot be accounted for by the phonon replica of the free exciton.In particular,it does not exhibit the characteristic asymmetric shape.We thus have to consider an electronic defect or an impurity as the possible origin of this band ͑real A band ͒.This is all the morelikely as the band shape can be accurately fit by a Gaussian curve,typical of defect or impurity contributions ͑see Fig.5͒.However,the NPs are made from the ablation of the mi-crocrystals of this sample.The ablation results from the in-teraction of the target with an intense pulsed laser ͑2ϫ106J per pulse of 10ns duration,focused on a 1mm 2spot,with a 10Hz repetition rate ͒.The instant energy deliv-ered to the target is so important that the ablation process produces a plasma whose composition is identical to the tar-get one ͑no selective ablation ͒.37Furthermore,the LECBD is a kinetically ruled process,far from the thermodynamic equi-librium.The nucleation of the clusters proceeds via an accre-tion process with an estimated cooling rate of 1010K /s.38At such a drastic cooling rate,the sticking coefficient of the species present in the plasma is close to one,39meaning that the clusters assembly has the same stoichiometry as the plasma.Only species that have extremely stable dimers ͑H 2,O 2,or N 2͒are likely to escape from the nucleation process.Even this last statement must be limited when considering weak impurity amounts.Since the accretion is a Poissonnian statistical process,the occurrence of an impurity dimer is very unlikely.Consequently,if an impurity were present in the initial ZnO powder,its luminescent signature would most probably be visible for both the ␮Cs and NPs samples.This is not the case.Therefore,a more likely hypothesis that can be invoked to explain the difference in the spectra of NPs and ␮Cs is a crystalline defect.In order to get more insight into the type of defect,we study the temperature dependence of the 3.31eV band ͑A band ͒in the ␮Cs spectra.The tem-perature dependence of the position of the defect induced band is generally the method used to differentiate the elec-tronic nature of the defect.The results are exposed in Fig.6.The three main mechanisms invoked as possible origins of the defect related band are an exciton bound to an accep-tor defect ͑AX ͒,a donor-acceptor pair recombination ͑DAP ͒and a FB transition.For the first two hypotheses the tempera-ture dependence of the transition follows the band-gap-energy variation according,respectively,to 6,31E n e r g y (e V )Temperature(K)FIG.4.͑Color online ͒Energies of the longitudinal phonon de-duced from the modified Permogorov law applied to emission spec-tra of the single crystal at different temperatures.The dashed area reproduces the range of suitable phonon energy according to the literature.I n t e n s i t y (a .u .)Energy (eV)3.27 3.30 3.33 3.36 3.39 3.42FXDXI n t e n s i t y (a .u .)Energy (eV)A bandT=80KFIG.5.͑Color online ͒Temperature dependence of the microc-rystals excitonic emission.The two dashed lines show examples of fit accuracy for the A band by a Gaussian distribution at 10and 80K.E n e r g y (e V )Temperature (K)FIG.6.͑Color online ͒Temperature dependence of the FX en-ergy ͑triangles ͒and the corresponding fit using the Manoogian law ͑solid line ͒.Temperature dependence of the difference in energy between the FX and the A band energies ͑circles ͒.The variation can be fitted by a linear law,indicative of a free-to-bound transition.COMPETITION BETWEEN EXCITON-PHONON …PHYSICAL REVIEW B 81,115304͑2010͒E AX͑T͒=E g͑T͒−E b−E loc a,͑3͒E DAP͑T͒=E g͑T͒+E c−E a−E d,͑4͒where E AX is the energy of the acceptor-bound exciton emis-sion,E DAP the energy of the donor-acceptor pair recombina-tion,E g the band-gap energy,E b the binding energy of the exciton,E a͑resp.E d͒the energy of the acceptor͑resp.donor͒level,E c the energy of the Coulomb interaction between thedonor and the acceptor,and E loc a the localization energy of the exciton on the acceptor center.Among all these param-eters,only E g depends significantly on the temperature,31 which implies that the energy dependence of these two tran-sitions follows the FX one.For the temperature dependence of the free-to-bound transition,we have to add a term which models the temperature dependence of the free-electron den-sity of state.Surprisingly,different temperature dependencies have been evoked in the literature for the same mechanism.4,40,41The stone mark work of Colbow40is well suited for wurtzite-type polar semiconductors and is thus the most justified model in our case.Assuming a parabolic con-duction band and a Boltzmann distribution for the free elec-trons,Colbow writes the temperature dependence of the tran-sition as follows:E FB͑T͒=E g͑T͒−E a+␣kT͑5͒with␣=1.In our case we have determined the temperature dependence of the band-gap energy byfitting the FX tem-perature dependence with a Manoogian law42as can be seen in Fig.6.This law is more adapted to the present study than the well-known Varshni law,43which is originally demonstrated for cubic materials and is not adapted forfit-ting the band-gap temperature dependence on a small tem-perature range.The parameters found here are close to the ones determined by Hamby et al.44on a ZnO single crystal and their discussion is out of scope of the present study. We then have plotted the temperature dependence of the dif-ference E FX−E A band in Fig.6.The result,over a130K temperature range,clearly shows a linear dependence with ␣=0.97Ϯ0.04.Since the free-exciton energy follows the gap variation,this confirms the assignment of the band to a free-to-bound transition.Schirra et al.45have shown that this tran-sition occurs near extended stacking faults in the basal plane. The absence of the3.31eV band in the nanoparticles and in the single-crystal spectra at low temperature is in accordance with this hypothesis.Indeed,these stacking faults consist in a change in the crystal symmetry between the wurtzite phase and the zinc blende one,a phenomenon common to hexago-nal semiconductors from SiC͑Ref.46͒to CdS.47They are thus well-defined and extended defects͑several nanometers wide͒.Their presence is unlikely both in an individual cluster of a few nanometers in diameter and in the assembly of such clusters.If not so,it would require some matching of the basal planes between clusters,inducing stacking faults, which we have not observed on a large scale by TEM.Even-tually,regarding the single crystal,no large amount of stak-ing faults of the basal plane is to be expected in such a highly crystallized sample.Furthermore,knowing the binding en-ergy of the free exciton͑ϳ60meV͒,it is possible to deduce from the temperature dependence of the quantity E FX −E A band the value of the activation energy of the related defect.We obtain the value of122Ϯ5meV,which is in good accordance with the value of130meV reported by Schirra et al.45D.Implication for the optical signature of p dopingThe possibility to use the3.31eV emission as an univocal signature of the p doping in ZnO structures,and particularly, in nanostructures,is a long-standing issue.As a result of the present study,on top of recent results from the literature,45,48,49it appears that this concept is to be used with great caution.Since the3.31eV band is partly related to the phonon replica of the FX,the characterization of p doping should befirst carried out at temperatures below80K or so. Second,the absence of any A band͑at low temperature͒in undoped material should be demonstrated prior to the dop-ing.This would ensure that no defect contributes to the band. Studies from the literature illustrate the fact that the3.31eV emission magnitude highly depends on the quality of the sample synthesis.For instance,ZnO nanopillars,with no im-purity intentionally incorporated,either grown on sapphire or silicon substrates exhibit an emission at3.31eV in the latter case48while they do not in the former case.49This observa-tion is to be related to the crystalline quality of the samples, which is better in the case of nanopillars grown on sapphire. To further support this,we can mention the absence of the 3.31eV emission in highly faceted ZnO rods,well crystal-lized in the wurtzite structure.50It may be argued that,even if the3.31eV emission is related to crystalline defects,since these may be of acceptor nature,they naturally contribute to the compensation of the n-doping background.Hence,the3.31eV band would still be a sound signature of p doping.Unfortunately,it seems that the issue is not that simple since scanning capacitance measurements on ZnO layers grown by pulsed laser deposi-tion have shown that regions of p-type conductivity coexist with regions of n-type conductivity among the same sample.51The former regions contain more defects than the latter.This observation tends to limit the role of defects as compensating centers.That is the reason why the ability to synthesize intrinsic samples is a crucial step before achieving p doping.When intentional p doping is aimed at,the incorporation of dopants as nitrogen,7phosphorus,5,52,53or arsenic54can be accompanied with the presence of the3.31eV emission. Therefore,the link between the incorporation of these ele-ments and the3.31eV emission is established.Schirra et al.,45in their critical review of the literature regarding p doping in ZnO,suggest that the previous dopants may induce the formation of crystalline defects͑stacking fault of the basal plane͒which leads to the appearance of the3.31eV emission but not necessarily to a good p-type conductivity because of a low-carrier mobility and strong spatialfluctua-tions of the doping levels.In this context,our synthesis technique to prepare ZnO nanoparticles is of great interest.They do not exhibit any defect related emission at3.31eV.Since their size and theirTAINOFF et al.PHYSICAL REVIEW B81,115304͑2010͒。