Cathodoluminescence in a Scanning Transmission ElectronMicroscope: A Nanometer-Scale Counterpart of Photoluminescencefor the Study of II−VI Quantum DotsZackaria Mahfoud,† Arjen T. Dijksman,¶ Clementine Javaux,¶ Pierre Bassoul,¶ Anne-Laure Baudrion,‡

Jero me Plain,‡ Benoît Dubertret,¶ and Mathieu Kociak*,†

†Laboratoire de Physique des Solides, CNRS/UMR8502, Universite Paris-Sud, Batiment 510, Orsay 91405, France‡Laboratoire de Nanotechnologie et d’Instrumentation Optique, Institut Charles Delaunay, UMR CNRS 6279, Universite deTechnologie de Troyes, France¶Laboratoire de Physique et d’Etude des Materiaux, UMR 8213, Ecole Superieure de Physique et de Chimie Industrielles, ParisTech,10 rue Vauquelin, 75005 Paris, France

ABSTRACT: We report on nanometer-scale cathodoluminescence (nanoCL) experi-ments in a scanning transmission electron microscope on individual core−shell CdSe/CdS quantum dots (QDs). By performing combined photoluminescence (PL) andnanoCL experiments of the same individual QDs, we first show that bothspectroscopies can be used equally well to probe the spectral properties of QDs. Wethen demonstrate that the spatial resolution of the nanoCL is only limited by the sizeof the QDs themselves by performing nanoCL experiments on QDs lying side by side.Finally, we show how nanoCL can be advantageous with respect to PL as it can rapidlyand efficiently characterize the optical properties of a large set of individual QDs. Theseresults contrast with pioneering CL works on II−VI QDs and pave the way to thecharacterization of any II−VI quantum-confined structure at the relevant scale.

SECTION: Plasmonics, Optical Materials, and Hard Matter

The study of quantum-confined structures, in particular,those composed of II−VI compounds, has exploded in

the past 20 years. Indeed, the advances in synthesis of quantumwells,1 quantum rods,2 and quantum dots (QDs)3 make itpossible now to manufacture 100% quantum efficiency,nonblinking QDs.4 The long predicted uses of such QDs,from biomarkers to quantum cryptography, are now at hand.Such objects are, for obvious physical reasons, much smaller

than the optical diffraction limit. The mandatory opticalinvestigations, usually done via absorption or photolumines-cence (PL) spectroscopies are thus limited to sets of QDs orwell-separated individual QDs.5 This makes it impossible toretrieve individual QD optical properties in a compactenvironment or to quickly assess the optical properties of alarge set of QDs and disentangle, for example, homogeneousand inhomogenous broadenings. Such an analysis is howevercrucial for the study of real devices.6

Alternatives to purely optical methods are thus needed. In apioneering work,7 Erni et al. have used electron energy lossspectroscopy (EELS) in a scanning transmission electronmicroscope (STEM). Despite the impressive proved benefit ofthe nanometric spatial resolution offered by electron-basedspectroscopies, fundamental and technical issues led unfortu-nately to inadequate spectral resolutions. Owing to itspotentially similar high spatial resolution but much betterspectral resolution, cathodoluminescence (CL) in an electron

microscope, in particular, in scanning EM (SEM-CL), should inprinciple be an excellent candidate.Indeed, individual III−V QDs have been successfully studied

by means of SEM-CL.8−11 However, the spatial resolution waslarger than the QD size by approximatively an order ofmagnitude. The reasons for the loss of spatial resolution arenumerous11 but are generally related to the very stronginteraction of the electron beam with the sample at therelatively small accelerating voltages used in SEM-CL (seebelow). This precludes the extension of the technique to thestudy of packed sets of QDs or to individual QD internalstructures. It is worth stressing also that in these works,8−10 therelationship between CL and PL spectra was not studied.The situation for II−VI QDs is even worse. First, it is worth

noting that the higher ionicity of the bonds in II−VIcompounds makes them much more prone to damage throughelectron beam irradiation than, for example, III−V compounds.However, even discarding this issue that is predominatelyimportant when using high current densities such as needed forgetting nanometer resolutions, other limitations exist. Indeed,in their pioneering works, Rodriguez-Viejo et al.12 performed acomparison of the optical properties of films of CdSe/ZnS QDsas probed by PL and CL in a SEM. The CL was performed

Received: October 16, 2013Accepted: November 15, 2013

Letter

pubs.acs.org/JPCL

© XXXX American Chemical Society 4090 dx.doi.org/10.1021/jz402233x | J. Phys. Chem. Lett. 2013, 4, 4090−4094

without spatial resolution. Although spectra were dominated bya single peak in both cases, major differences were observed.The CL peak was red-shifted by around 6 nm, and the peak fullwidth at half-maximum (fwhm) was much broader (40−50 ascompared to 30−40 nm for the PL). The former effect wasattributed to nonlinearities, possibly related to the quantumconfined Stark effect (QCSE)13 and the latter to the effect ofelectron beam heating. Quenching of the emission wasobserved over time. Finally, a huge secondary electron emission(SEE) triggering further delocalized CL emission wasdiagnosed. All of these results, the absence of spatial resolution,nonlinearities, heating, SEE, and so forth, were related to thehuge interaction of the electron beam with the sample in a SEMand the high electron beam current used.In this Letter, we will present combined PL and STEM-CL

(nanoCL) measurements of spectra from individual core−shellCdSe/CdS QDs. We will show that the PL and CL spectra areequivalent and thus that nanoCL can be used as a nanometer-scale counterpart to PL for the study of II−VI QDs. We willthen present nanometer-scale nanoCL experiments showinghow individual QD optical properties can be obtained evenwhen they are close-packed and how a very large number ofQDs can be investigated at once.PL experiments were performed at room temperature with a

confocal microscope (MicroTime, PicoQuant) and a Han-bury−Brown and Twiss setup based on two avalanchephotodiodes and a single photon counting module (HydraHarp400, Picoquant). The individual QDs were excited with apulsed diode emitting at 402 nm, and the emitted light wascollected with an air objective (100×, NA = 0.9). All or part ofthe emitted light was directed toward a spectrometer with a 150l/mm grating and a resolution of ∼0.8 meV (Shemrock 750,Andor Technology).CL experiments were performed at 170 ± 20 K in a VG HB

501 STEM, fitted with a homemade nanoCL system.14 Informer works, this setup allowed us to disentangle the spectralsignatures of individual III−V, 1−5 nm thick, quantum wellsseparated by less than 4 nm,14 with a minimum QCSE. Also,thanks to a high-efficiency design that collects most of thecathodo-generated photons and thus minimizes the electron

beam current, single-photon detection from nitrogen vacancycenters in nanodiamond could be observed,15 implying thepossibility of addressing individual II−VI QDs in a linearregime with the same setup.In the present experiments, typical conditions were a probe

size around 1 nm, accelerating voltage of 60 keV, and a currentaround 1 nA. Note that accelerating voltages from 40 to 100keV were also used without detectable differences in the results.Experiments were performed in the spectral imaging mode. Inthis mode, the beam is scanned over the region of interest. Ateach scan position, a spectrum is recorded, along with the high-angle annular dark field (HAADF) signal. At the end of thescan, both a CL spectrum image (SI), that is, an image with aspectrum per pixel, and a HAADF image are available. Fromthe latter, the local thickness can be determined. In thefollowing, spectra and maps are directly extracted from such aSI.We have used CdSe/CdS QDs with a core between 3 and 5

nm in diameter and with a relatively thick shell (10 nm). TheseQDs have been shown to have a bright emission and almost noblinking.16 Such optimized QDs were important for the successof the experiments, which were dedicated to study the spectralproperties of the QDs and to perform combined PL/CLmeasurements. Spectral measurements usually require longexposure times and high signal levels because the detection isparallel and one has to overcome losses in the detection path(imperfect CCD camera quantum efficiency, lenses andgratings losses, etc.). Detection schemes using a simplifiedoptical path with fast, high-sensitivity serial detectors (such asavalanche photodiodes) could be used to study less robustQDs. Indeed, we performed experiments on commercial QDs(Evident Technology) with thin shells. Although spectra couldbe obtained from large aggregates of QDs, individual QDsignals could not be obtained due to a very fast vanishing of thesignal; their study would benefit from a faster detection setup.The QDs were deposited on commercial Si or Si3N4

membranes that were thin enough (15 nm) to be electron-transparent (see Figure 1a). In the case of correlated PL/CLmeasurements, before deposition, the membrane was patternedby e-beam lithography with reference crosses that can be

Figure 1. Comparison between PL and CL spectra of individual CdSe/CdS QDs. (a) Scheme of the PL- and nanoCL-compatible substrate (all madeup of Si for the presented example). (b) HAADF image of the QD, the PL/CL spectroscopies of which are shown in (c). (c) CL and PL spectra ofthe same QD. The integration time is 1 min for PL and 50 ms for CL. The energy shift can be explained by a temperature difference between themeasurements (see text). (Inset, left) Scheme of the nanoCL/PL experiment. (right) PL time correlation curves for the same QD.

The Journal of Physical Chemistry Letters Letter

dx.doi.org/10.1021/jz402233x | J. Phys. Chem. Lett. 2013, 4, 4090−40944091

detected both in the photon and electron microscopes;17 clearlyseparated individual QDs can thus be localized and observed byboth techniques. Note that if the QDs are exposed to air for toolong (i.e., few days), both PL and CL signals vanish. ComparedCL/PL experiments were thus performed within the sameweek. Numerous QDs were studied, hundreds by individual PLor nanoCL experiments and more than 20 in correlated CL/PLexperiments. In one case, we performed a PL/CL/PL seriesshowing no spectral shift between both PL experiments but adecrease in lifetime. This implies a specific, but minor,deleterious effect of the e-beam on the optical properties.However, in both PL and CL, excessively long observations ledto a bleaching of the QDs. In most cases, the bleaching in CL isassociated with a morphological change in the QDs.Figure 1 presents the comparison between a PL and a CL

experiment performed on the same QD. The HAADF (Figure1b) exhibits the typical diamond-like shape of the QDs. Thefact that the observed object is truly a QD is ascertained by thePL time correlation function showing a zero delay peak smallerthan 0.5 (inset of Figure 1c). In the three investigated QDsshowing CL and PL emissions, CL and PL have very similarspectra, made up of a single peak at around 620 nm, as expectedgiven the 3 nm core diameter. The CL is howeversystematically blue-shifted by 10−16 nm with respect to thePL. The CL peak fwhms, on the order of 14 ± 5 nm (around28 meV), are also systematically smaller than those in PL by afactor of 1.2−1.5. The former effect is well-explained by thedifference in temperature between the experiments; a blue shiftof around 14 nm is expected between 300 and 170 K.18 Thelatter is most probably related to the spectral diffusion19 as the

fwhms in both spectroscopies are already much larger than thetemperature spreading. In the PL experiment, the integrationtime was 1 min, and that of the CL experiments was only 50ms, which would explain the larger broadening for the formerspectroscopy.This clearly shows that nanoCL in a STEM and PL provides

almost the same spectral information on individual QDs,contrary to the results of previous SEM-CL experiments,12 witha minimal introduction of nonlinear and heating effects due tothe electron beam.Figure 2 presents SI mapping of a set of six QDs packed

together. Again, the diamond-like shape of the QDs is clearlyvisible in the HAADF image (Figure 2a). The HAADF takenalong with a SI is shown in Figure 2b. In this case, the emissionspectrum taken on different QDs (Figure 2c) is almost thesame, and no variation of spectral shape is seen within a singleQD. Also, the fact that the same QD can be probed severaltimes during the scan is a good indication that the nanoCL isnot inducing rapid damages and cannot be responsible for theabsence of emission of the neighboring nonemitting QDs. Bycomparing the HAADF image (Figure 2b) and the intensitymap extracted from the SI acquired in parallel (Figure 2d), onenotes that within individual QDs, the intensity increasesroughly monotonically with the local thickness. More strikingly,it is obvious that only three QDs are emitting. Such informationwould have been unavailable from a PL experiment.Interestingly enough, no correlation with a specific feature inthe QD (orientation, presence of defects, and so forth) isobserved here or in any other nanoCL experiment to explainthe presence or absence of emission.

Figure 2. High spatial resolution CL mapping of the QDs. (a) HAADF image of a set of six QDs. (b) HAADF of the same set acquired in parallel toa SI. (c) Spectra extracted from the SI on areas indicated on (b). (d) CL intensity map. Only three QDs are emitting. The scale bars in (c) and (d)are 20 nm.

The Journal of Physical Chemistry Letters Letter

dx.doi.org/10.1021/jz402233x | J. Phys. Chem. Lett. 2013, 4, 4090−40944092

This example allows us to set a limit in the spatial resolution.Usually, four main parameters affect the spatial resolution of aCL experiment. First is the size of the electron probe, which ishere typically less than 1 nm and thus not relevant. Second isthe SEE, triggering photon emission from the neighborhood ofthe impact area.12 This effect is here clearly ruled out by theabsence of detected CL signal when the beam is hitting the darkQDs. If the SEE were retriggering photon emission in theneighborhood, an emission would have been measured at thesepoints. Third is the broadening of the emission location due tomultiple deflections and interactions of the incident electronbefore it is absorbed. This broadening is known as the“excitation pear”. Here, the effect is shown to be negligible as avariation of the emitted intensity within individual QDs can bedetected (see Figure 2d). This rules out the existence of a broadexcitation pear. The absence of an excitation pear is indeedexpected for a 30 (QD) + 15 (substrate) nm thick sample at 60keV. Fourth is that the charge carriers created by the incomingelectron at a given point may diffuse in the sample andrecombine far away from the point where they have beencreated. This effect, known as “charge carrier diffusion”, mightagain degrade the spatial resolution as the emitted light mightcome from a place far from the excitation point. As mentionedabove, the signal intensity within a QD seems to scale with itsthickness. No wavelength or fwhm changes can be detectedwithin an individual QD. From the point of view of theelectron/matter interaction, this means that either the diffusionlength is larger than the QD size or the QD responds as awhole to the electronic excitation. Either way, the absence ofmeasurable spectroscopic change within an individual QDmeans that the effective spatial resolution is set by the QD sizeitself.We now show how it is possible to explore the statistics in

the samples by analyzing large SI.

Figure 3a presents a HAADF image of a large set of QDs, inwhich the emission changes from point to point (Figure 3b).On this set, we have acquired a SI in which the spatial samplinghas been set approximatively to the size of a single QD, as seenon the HAADF image acquired in parallel (Figure 3c). Theacquisition time of the SI is around 10 min. As in the previouscase, comparing the HAADF and the SI results, we observe thatsome of the QDs do not emit, as clearly seen on the intensitymap Figure 3d. Also, the intensity variations follow thethickness variations. The wavelength dispersion from QD toQD (Figure 3b and e) is on the order of 40 nm, and nocorrelation could be found between the wavelength changesand the thickness variations. The fwhm spatial variations shownin Figure 3f do not follow any clear trend. From the fact thatneither the wavelength nor the fwhm follows the thickness, wecan deduce that no obvious heating effects are seen in nanoCL,even when several QDs may be stacked along the electronbeam direction. The number of pixels containing at least oneQD is larger than 2000, with approximatively 600 that areemitting. This shows how quickly information on the opticalproperties of a dense ensemble of QDs can be obtained. Thefact that a large number of QDs are not emitting is related tothe imperfect yield of this new synthesis of thick-shell QDs,leading to the likely presence of nonradiative centers at thesurface of the QDs.We have shown that nanoCL spectroscopy can be used to

study individual II−VI QD optical properties. By comparingwith PL spectroscopy from the very same individual QDs, wehave shown that both spectroscopies offer essentially the samespectral information. Common sources of heating andnonlinearities previously attributed to electronic excitation12

are here shown to be irrelevant in a STEM. While PL is limitedby the optical diffraction limit and regular (low-voltage) SEMon thick samples is limited by the excitation pear and SE

Figure 3. Batch measurement on a large set of QDs. (a) HAADF image of the set of QDs. Individual QDs can be seen. (b) Selected spectra extractedfrom areas circled on (a). (c) HAADF image acquired during a SI on the set. (d) Intensity; (e) wavelength; (f) fwhm maps extracted from the SI.Scale bars in (c−f) are 200 nm.

The Journal of Physical Chemistry Letters Letter

dx.doi.org/10.1021/jz402233x | J. Phys. Chem. Lett. 2013, 4, 4090−40944093

reemission, nanoCL is only limited by the object size, whichitself defines the necessary spatial resolution. We have shownhow this gain in resolution can be used to rapidly characterizethe emission properties of a given set of individual QDs. Wefocused in this work on the spectral information, which requiresrelatively long acquisition times. It might be interesting toupgrade the setup with fast, high-sensitivity, low-loss serialdetection. This would allow us to address with the improvedspatial resolution other issues like QD blinking.20 This workpaves the way toward the systematic study of II−VI and III−VQDs at the single QD scale. Also, it offers the possibility ofstudying plasmon/II−VI QD coupling at the nanometer scale,much in the same way as has been done for III−V systems.10

■ AUTHOR INFORMATION

Corresponding Author*E-mail: kociak@lps.u-psud.fr.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

M.K. thanks M. G. Walls for his careful reading of themanuscript. This work has received support from the FrenchNational Agency for Research under the program of futureinvestment TEMPOS-CHROMATEM with the referenceANR-10-EQPX-50 and the HYNNA project. The researchleading to these results has received funding from the EuropeanUnion Seventh Framework Programme [No. FP7/2007- 2013]under Grant Agreement No. n312483 (ESTEEM2). Theauthors acknowledge the Region Champagne-Ardennes, theConseil general de l’Aube, and the FEDER funds through theirsupport of the regional platform Nano’mat.

■ REFERENCES(1) Ithurria, S.; Dubertret, B. Quasi 2D Colloidal CdSe Platelets withThicknesses Controlled at the Atomic Level. J. Am. Chem. Soc. 2008,130, 16504.(2) Peng, X. G.; Manna, L.; Yang, W. D.; Wickham, J.; Scher, E.;Kadavanich, A.; Alivisatos, A. P. Shape Control of CdSe Nanocrystals.Nature 2000, 404, 59−61.(3) Garcia-Rodriguez, R.; Hendricks, M. P.; Cossairt, B. M.; Liu, H.;Owen, J. S. Conversion Reactions of Cadmium ChalcogenideNanocrystal Precursors. Chem. Mater. 2013, 25, 1233−1249.(4) Mahler, B.; Spinicelli, P.; Buil, S.; Quelin, X.; Hermier, J. P.;Dubertret, B. Towards Nonblinking Colloidal Quantum Dots. Nat.Mater. 2008, 7, 659−664.(5) Plain, J.; Sonnefraud, Y.; Viste, P.; Lerondel, G.; Huant, S.; Royer,P. Self-Assembly Drives Quantum Dot Photoluminescence. J. Fluoresc.2009, 19, 311−316.(6) Kim, T. H.; Cho, K. S.; Lee, E. K.; Lee, S. J.; Chae, J.; Kim, J. W.;Kim, D. H.; Kwon, J. Y.; Amaratunga, G.; Lee, S. Y.; Choi, B. L.; Kuk,Y.; Kim, J. M.; Kim, K. Full-Colour Quantum Dot Displays Fabricatedby Transfer Printing. Nat. Photonics 2011, 5, 176−182.(7) Erni, R.; Browning, N. D. Quantification of the Size-DependentEnergy Gap of Individual CdSe Quantum Dots by Valence ElectronEnergy-Loss Spectroscopy. Ultramicroscopy 2007, 107, 267−273.(8) Grundmann, M.; Christen, J.; Ledentsov, N. N.; Bohrer, J.;Bimberg, D.; Ruvimov, S. S.; Werner, P.; Richter, U.; Gosele, U.;Heydenreich, J.; Ustinov, V. M.; Egorov, A. Y.; Zhukov, A. E.; Kopev,P. S.; Alferov, Z. I. Ultranarrow Luminescence Lines From SingleQuantum Dots. Phys. Rev. Lett. 1995, 74, 4043−4046.(9) Garayt, J. P.; Gerard, J. M.; Enjalbert, F.; Ferlazzo, L.; Founta, S.;Martinez-Guerrero, E.; Rol, F.; Araujo, D.; Cox, R.; Daudin, B.; Gayral,B.; Dang, L. S.; Mariette, H. Study of Isolated Cubic GaN Quantum

Dots by Low-Temperature Cathodoluminescence. Phys. E 2005, 26,203−206.(10) Nogues, G.; Merotto, Q.; Bachelier, G.; Lee, E. H.; Dong Song,J. Fabrication and Tuning of Plasmonic Optical Nanoantennas aroundDroplet Epitaxy Quantum Dots by Cathodoluminescence. App. Phys.Lett. 2013, 102, 231112.(11) Yacobi, B.; Holt, D. Cathodoluminescence Microscopy of InorganicSolids; Springer: New York, 1990.(12) Rodriguez-Viejo, J.; Jensen, K. F.; Mattoussi, H.; Michel, J.;Dabbousi, B. O.; Bawendi, M. G. Cathodoluminescence andPhotoluminescence of Highly Luminescent CdSe/ZnS QuantumDot Composites. Appl. Phys. Lett. 1997, 70, 2132−2134.(13) Miller, D. A. B.; Chemla, D. S.; Damen, T. C.; Gossard, A. C.;Wiegmann, W.; Wood, T. H.; Burrus, C. A. Band-Edge Electro-absorption In Quantum Well Structures The Quantum- ConfinedStark-Effect. Phys. Rev. Lett. 1984, 53, 2173−2176.(14) Zagonel, L. F.; Mazzucco, S.; Tence, M.; March, K.; Bernard, R.;Laslier, B.; Jacopin, G.; Tchernycheva, M.; Rigutti, L.; Julien, F. H.;Songmuang, R.; Kociak, M. Nanometer Scale Spectral Imaging ofQuantum Emitters in Nanowires and Its Correlation to TheirAtomically Resolved Structure. Nano Lett 2011, 11, 568−573.(15) Tizei, L. H. G.; Kociak, M. Spatially Resolved Quantum Nano-Optics of Single Photons Using an Electron Microscope. Phys. Rev.Lett. 2013, 110, 153604.(16) Javaux, C.; Mahler, B.; Dubertret, B.; Shabaev, A.; Rodina, A. V.;Efros, A. L.; Yakovlev, D. R.; Liu, F.; Bayer, M.; Camps, G.; Biadala, L.;Buil, S.; Quelin, X.; Hermier, J. P. Thermal Activation of Non-Radiative Auger Recombination in Charged Colloidal Nanocrystals.Nat. Nanotechnol. 2013, 8, 206−212.(17) Koberling, F.; Mews, A.; Philipp, G.; Kolb, U.; Potapova, I.;Burghard, M.; Basche, T. Fluorescence Spectroscopy and Trans-mission Electron Microscopy of the Same Isolated SemiconductorNanocrystals. Appl. Phys. Lett. 2002, 81, 1116−1118.(18) Al Salman, A.; Tortschanoff, A.; Mohamed, M. B.; Tonti, D.;van Mourik, F.; Chergui, M. Temperature Effects on the SpectralProperties of Colloidal CdSe Nanodots, Nanorods, And Tetrapods.Appl. Phys. Lett. 2007, 90, 093104.(19) Fernee, M. J.; Plakhotnik, T.; Louyer, Y.; Littleton, B. N.;Potzner, C.; Tamarat, P.; Mulvaney, P.; Lounis, B. SpontaneousSpectral Diffusion in CdSe Quantum Dots. J. Phys. Chem. Lett. 2012, 3,1716−1720.(20) Spinicelli, P.; Buil, S.; Quelin, X.; Mahler, B.; Dubertret, B.;Hermier, J. P. Bright and Grey States in CdSe−CdS NanocrystalsExhibiting Strongly Reduced Blinking. Phys. Rev. Lett. 2009, 102,136801.

The Journal of Physical Chemistry Letters Letter

dx.doi.org/10.1021/jz402233x | J. Phys. Chem. Lett. 2013, 4, 4090−40944094