murray, norris y bawendi

8/12/2019 Murray, Norris y Bawendi

1/10

8706 J . Am . Chem. Soc. 1993, 115, 8706-8715

Synthesis and Characterization of Nearly Monodisperse CdE(E = S, Se, Te) Semiconductor NanocrystallitesC. B. Murray, D. J. Noms, and M. G. BawendiContribution fr om the Departmen t of Che mistry, Massachusetts Institute of Technolog y,Camb ridge, Massachusetts 02139Received March 22, 1993

Abstract: A simple route to th e production of high-quality CdE (E = S, Se, Te) sem icondu ctor nanocrystallites ispresented. Crysta llites from - 2 A to - 15 A in diame ter with consistent crystal structure, surface derivatization,and a high degree of monodispersity are prepared in a single reaction . Th e synthesis is based on the pyrolysis oforgan ome tallic reagen ts by injection into a hot coord inating solvent. This provides tempora lly discrete nucleation a ndperm its controlled grow th of macroscopic quantities of nanocrystallites. Size selective precipita tion of crystallites fromportions of the growth solution isolates samples with narrow size distribu tions (


2/10

Synthesis of Cd E Semiconductor Nanocrystallitesin sufficient TO P to produce 1.OM stock solutions of trioctylphosph ineselenide [TOPSe] an d trioctylphosphine telluride [TOPT e] 5Method 1. The typical preparation of TOP/TOPO capped CdSenanocrystallites follows: Fifty grams of T OP O is dried and degassed inthe reaction vessel by heating to -200 C at -1 Torr for -20 min,flushing periodically with argon. The tem perature of the reaction flaskis then stabilized at -300 C under - atm of argon.Solution A is prepared by adding 1 OO mL (1 3.35 mmol) of M ezCdto 25.0 m L of TO P in the drybox. Solution B is prepared by adding 10.0mL of the 1 O M T OPSe stock solution (10.00 mmol) to 15.0 mL of TO P.Solutions A and B are com bined and loaded into a 50-mL syringe in thedrybox.Th e heat is removed from the reaction vessel. The syringe containingthe reagent m ixture is quickly removed from the drybox and its contentsdelivered to the vigorously stirring reaction flask in a single injectionthrough a rubber septum. Th e rapid introduction of the reagent mixtureproduces a deep yellow/orange solution with an absorption feature at440-460 nm. This is also accompanied by a sudden decrease intemperature to - 80 O C . Heating is restored to the reaction flask andthe temperature is gradually raised to 230-260 C.Aliquots of the reaction solution are removed at regular intervals ( 5 -10 min) and absorption spectra taken to m onitor the growth of thecrystallites. Th e best quality sam ples are prepared over a period of a fewhours of steady growth by modulating the grow th temp erature in responseto changes in the size distribution as estimated from the absorption spectra.The tem peratu re is lowered in response to a spreading of the sizedistribution and increased when growth appears to stop. When the desiredabsorption characteristics are observed, a portion of the growth solutionis transferred by cannula an d stored in a vial. In this way, a series of sizesranging from - 5 to 115 8, in diameter can be isolated from a singlepreparation.CdTe nanocrystallites are prepared by Method 1 with TOP Te as thechalcogen source, an injection temperature of -240 O C and growthtemperatures between -190 and -220 O C .Method 2. A second route to the production of CdE (E = S,Se, Te)nanocrystallites replaces the phosph ine chalconide precursors in Method1 with (TMS)zS, (TMS)zSe, and (BDMS)zTe, respectively. Growthtemperatures between -290 and -320 O C were found to provide thebest Cd S samples. T he smallest - 12 A) CdS, CdSe, and CdTe speciesare produced under milder conditions with injection and growth carriedout a t -100 O C .Isolation and Purification of Crystallites. A IO-mL aliquot of thereaction solution is removed by cannula a nd cooled to -60 O C slightlyabove the melting point of TOPO . Addition of 20 mL of anhydro usmethano l to the a liquot results in the reversible flocculation of thenanocrystallites. The flocculate is separated from the supernatant bycentrifugation. Dispersion of the flocculate in 25 mL of anhydro us1-butanol followed by further c entrifugation re sults in an optically clearsolution of nanocrysta llites and a gray precipitate containin g byproductsof the reaction. Powder X-ray diffraction and energy dispersive X-rayanalysis indicate these byproducts consist mostly of elem ental Cd and S e.This precipitate is discarded. Addition of 25 mL of anhydrous methanolto the sup ernatan t produces flocculation of the crystallites and removesexcess TO P and TO PO. A final rinse of the flocculate with 50 mL ofmethanol and subsequent vacuum drying produces -300 mg of freeflowing TO P/TO PO capped CdSenanocrystallites. The resulting powderis readily dispersed in a variety of alkane s, aromatics, long-chain alcohols,chlorinated so lvents, and org anic bases (amines, pyridines, furans,phosphines).Size-Selective Precipitation. Purified nanocrystallites are dispersedin anhydro us 1-butano l forming an optically clear solution. Anhyd rousmethano l is then add ed dropwise o thedispersio n until opalescence persistsupon stirring or sonication. Separation of supern atant and floccu late bycentrifugation produc es a precipitate enriched with the largest crystallitesin the sample. Dispersionof the p recipitate in 1-butanol and size-selectiveprecipitation with methano l is repeated until no further sharpening of theoptical absorption spectrum is noted. Size-selectiveprecipitation can becarried out in a variety of solvent/non solvent pairs, including pyridine/hexane and chloroform/methanol.Surface Exchange. Crystallite surface derivatization can be m odifiedby repeated exposure o an excess of a competing capping group. Heat ingto -60 OC mixture of -50 mg of TO PO /TO P capped crystallites and5-10 mL of pyridine gradually dispe rses the crystallites in the solvent.

( 5 ) Zingaro, R. A,;Steeves, B.H ;rgorlic, K. J.Orgunomet. Chem. 1965,4 , 320.

J . A m . Chem. SOC.,Vol. 115, No. 19, 1993 8707Treatm ent of the dispersion with excess hexane results in the flocculationof the crystallites which are then isolated by centrifugation. Th e processof dispersion in pyridine and flo cculationwith hexane is repeated a numberof times o producecrystalliteswhich disperse readily n pyridine, methan ol,and aromatics but no longer disperse in aliphatics.Optical Characterization. Optical absorption spectra were collectedat room temperatureon a Hewlett-Packard 8452 diode array spectrometerusing 1-cmquar tz cuvettes. Samples were prepared by dispersingwashedCdSe nanocrystallites n hexane. Luminescence xperimentswere carriedout on a S PE X Fluorolog-2 spectrometer with use of front face collectionwith500-wslits. Estimatesof quantum yield were obtained by com paringthe integrated emission from Rhodamine 640 in methanol and tha t of35 8 diameter CdSe nanocrys tallita dispersed in hexane. Concen trationsof both w ere adjusted to provide optical densities of 0.30 at 460 nm inmatched 1-mm quartz cuvettes. Fluorescence spectra were collectedbetween 480 and 800 nm a t room temperature with 460-nm excitation.Transmission Electron Microscopy. A Topcon EM002B electronmicrosco pe operating a t 200 kV was used for transmission electronmicroscopy (T EM ). Imaging was carried out in bright field with anobjective aperture selected to permit lattice imaging of the (loo ), (002),and (101) Wurtziteplanes. Copper grids (300 mesh) coated with a -50A amorphous carbon film were purchased from Ernest Fullam. Sampleswere prepared by placing a d rop of a dilute pyridine dispersion ofnanocrystallites on the surface of a grid, waiting for - min, and thenwicking away the solution. Th e coverage evel of crystallites was adjustedby varying the initial dispersion concentration an d the contact tim e.X-ray Powder Diffraction. Powder X-ray diffraction spectra werecollectedon a Rigaku 300 Rotaflex diffractometer opera ting in the Braggconfiguration using Cu Ka adiation. The accelerating voltage was setat 250 kV with a 200 milliamp flux. Scatte r and diffraction slits of 0.50and a 0.15-mm collection slit were used. Samples for X-ray diffractionwere prepared from - 5 0 0 mg of thoroughly washed and dried nano-crystallite powder. The free-flo wing powders were pressed a t S O 0 0 p ito form 0.5 in. diameter pellets with mirror flat surfaces.Hi. Results and Discussion

The produc t ion and phenomenologyof monodisperse lyophobiccolloids has been inves t iga ted s ince Faraday's produc t ion of goldsols in 1857.6 C l a s s i c w o r k by La Mer a n d D i n e g a r' has shownthat the produc t ion of a series of monodisperse lyophobic colloidsd e p e n d s on a t e m p o r a l l y d i s c r e t e n u c l e a t io n event followed bycont ro l led g rowth on t h e existing nuc le i . Temporally d i s c r e t enucleation in our syn thes i s is attained b y a r a p i d increase in ther e a g e n t c o n c e n t r a t i o n s upon injection, resulting in an abruptsuper sa tura t ion which i s re l i eved by the formation of nuclei a n dfollowed by growth on the in i t i a l ly formed nuclei.The w o r k of S t e i g e r w a l d a n d co-workers on the use ofo r g a n o m e t a l l i c precursors in the so lu t ion-phase syn thes i s of b u l kand nanocrystalline m a t e r i a l s p r o v i d es g u i d a n c e in our selectionof reagen ts .3a ,bJ MezCd is chosen a s the C d source a n d ( T M S ) z E(E = S, e,Te) or TOPSe a n d TOPTeare selectedas chalcogensources w i t h TOPSe a n d TOPTe prefe r red due to their ease ofp r e p a r at i o n a n d their stabil i ty. MezCd and ( T M S ) 2 E r e a g e n tsh a v e b e e n shown to undergo dea lky ls i ly la t ion in a variety ofsolvents as a route t o the produc t ion of b u l k materials.ET r i m e t h y l p h o s p h i n e t e l lu r i d e is k n o w n as a good source of TeO.9Mixed phosphine /phosphine ox ide solutions have previously beenf o u n d to b e good solvents for the h i g h - t e m p e r a t u r e g r o w t h anda n n e a l i n g of C d S e crystallites.2d,cJoT h e c o o r d i n a t in g solventplays a crucial ro le in cont ro l l ing the g r o w t h process, s tab i l i z ingthe resulting co l lo ida l d i spers ion , an d e lec t ron ica l ly passivatingt h e s e m i c o n d u c t or s u r f a c e .Nucleation and G r o w th . I n j e ct i o n o f r e a g e n t s i n t o t h e h o treac t ion pot resu l t s in a s h o r t b u r s t o f homogeneousnucleation.

(6) Overbeek, J. Th. G. Ado. Colloid Interfuce Sci. 1982, 15 251.(7)LaMer, V.K.; Dinenar, R. H. J. Am . Chem.Soc. 1950, 72. 4841.(8) Stuczvnski. S.M.: Brennan. J. G.; Steigerwald, M. L. Inora.Chem.1989,' 28, 443 1.(91 (al Steiaerwald. M. L.: Surinkle.C. R. J . Am. Chem.SOC. 987.109.. . . . .7200. b) Stergerwald, M . Chem. Mater. 1989, I 52.(10)Bawendi, M. G ; ortan, A. R.; Steigerwald, M. L.; Brus, L. E. .Chem. Phys. 1989, 91, 7282.


3/10

8108The depletion of reagents through nucleation and the suddentemperature drop associated with the introduction of roomtemperatu re reagents prevents further nucleation. Gently re-heating allows slow growth and annealing of the crystallites.Crystallite growth appears consistent with Ostwald ripening,where the higher surface free energy of small crystallites makesthem less stable with respect to dissolution in the solvent thanlarger crystallites. Th e net result of this stability gradient withina dispersion is slow diffusion of materia l from sm all particles tothe surface of larger partic1es.I Reiss has shown how growth bythis kind of transport can result in the production of highlymonodisperse colloidal dispersio ns rom systems hat may initiallybe polydisperse.I2Both the average size and the size distribution of crystallitesin a sample are dependent on the growth temperature, con sistentwith surface free energy considerations. The grow th temperaturenecessary to m aintain steady growth increases with increasingaverage crystallite size. As th e size distribution sharpens, thereaction temperature must be raised to maintain steady growth.Conversely, if the size distribution begins to spread, the tem-perature necessary for slow steady growth drops. Size distri-butions during growth are crudely estimated from absorptionline widths (typically 50nm fwhm). Modulation of the reactiontempera ture in response to changes in the absorption spectrumallows the maintenance of a sh arp size distribution as the samplegrows.The Ostwald ripening process accentuates any kinetic ortherm odyn amic bottleneck in the growth of crystallites. As abottleneck is approac hed (e.g. a closed structural shell), sharpeningof the sample size distribution reduces the therm odynamic drivingforce for further growth. Sharpening in the absorption featuresas the average particle size approaches 12 ,20 ,35 ,45 , and 51 Ain diameter m ay point to th e presence of such bottlenecks.Capping groups present a significant steric barrier to theadditionof mat erial to th e surface of a growing crystallite, slowingthe growth kinetics. The TO P/T OP O solvent coordinates thesurface of the crystallites and permits slow steady growth attemperatures above 280 OC. Replacing the octyl chains withshorter groups reduces the temperature for controlled growth.Mixed alkylphosphine/alkylphosphine xide solvents with bu tyl,ethyl, and methyl grou ps show uncontrolled growth at 230, 100,and 50 OC, respectively. Stead y controlled grow th resultsin highlymonodisperse particles of consistent crystal structure and allowssize selection by extrac ting sam ples periodically from the re actionvessel.A wealth of poten tialorganometallicprecursors and high boilingpoint coordinatin g solvents exist. Altho ugh phosphine/phosphineoxide have been found to provide the most controlled growthcond itions, injection s of reagents into hot pyridines, tertiar yamines, and furans all allow production of nano crystallites.We are beginning the extension of this synthetic method to theproduction of ZnE and HgE materials using diethylzinc anddibenzylmercury as group I1 sources. Growth conditions havenot yet been op timized to provide sample quality comparable tothat of CdE materials.Colloid Stabilization and Size-SelectivePrecipitation. Lyo-phobic colloidal particles attract each other by van der Waalsforces. The attraction is strong due to the near additivity offorces between pa irs of unit cells in differe nt particles.13 Colloidsremain stable with respect to aggregation only if there exists arepulsive force of sufficient strength and range to coun teract thevan der Waals attraction . Chemisorption of ambiphylic specieson the surface of the particles gives rise to a steric barrier toaggregation. Th e dispersions of CdS e nanocrystallites are

1 1 ) Smith,A. L.ParticleGr0wthinSuspensions;AcademicPress: ondon,1983; pp 3-15.(12) Reiss, H . Chem. Phys. 1951, 19, 482.(1 3) Sato, T.; Ruch, R. Stabilization of Colloidal Dispersions by Polymer

J . Am. Chem. SOC.,Vol 115, No. 19 1993

Absorption; Marcel Dekker: New York, 1980; pp 46-51.

Murray et al.

350 400 450 500 550 600Wavelength nm)

Figure 1. Example of the effect of size-selective precipitation on theabsorption spectrum of - 7A diameter CdSe nanocrystallites. (a)Roomtemperature optical absorption spectrum of the nanocrystallites in thegrowth solution before size-selective precipitation. (b) Spectrum afterone size-selective precipitation from the growth solution wit methanol.(c) Spectrum after dispersion in 1 butanol and size-selective precipitationwith methanol. (d) Spectrum after a final size-selective precipitationfrom I-butanol/methanol.sterically stabilized by a lyophylic coat of alky l group s anchoredto the crystallite surface by phosphine ox ide/cha lconide moieties.Th e efficiency of the steric stab ilization is strongly depend ent onthe interaction of the alkyl groups with the solvent. Gradua laddition of a nonsolvent can produce size-dependent locculationof the nano crystallite dispersion. This phenome non is exploitedin further narrowing the p article size distribution.The addition of methanol increases the average polarity of thesolvent and reduces the energetic barrier to flocculation. Th elargest particles in a dispersion experience the greatest attractiveforces. These large particles have a higher probab ility ofovercoming the reduced barrier and are thus enriched in thefloccu late produ ced. The remova l of a specific subset of particlesfrom th e initial size distribution narrows the size distribution ofboth supern atant and precipitate. Depending on the cap molecule,a number of solvent/nonsolvent systems can be used for size-selective precipitation (e.g. hexane/ethanol, chloroform/meth-anol, pyridinefhexane, etc.... .Figure 1 illustrates the result of size-selective precipitation.Spectrum a shows the optical absorption of the initial growthsolution. The broad ab sorption features correspond to a samplewith an average size of 35 A 10% (sized by TE M) . Slowaddition of methanol results in the flocculation of the largerparticles in the distribu tion which give spectrum b after dispersingin 1-bu tanol. Titration of methano l in samp le b again producesfloccula tion of the larger pa rticles, giving spectrum c up ondispersion in 1-buta nol. A final size-selective precipita tion from1-butanol yields a sample with optical absorption d and with anaverag e size of -3 7 A 5 . Spectrum d is dramaticallysharpe ned relative to that of th e initial growth solution and revealstransitions at 530 and -400 nm which were previously cloakedby polydispersity. For the fractionation process to work well itis crucially important tha t the shape and surface derivatizationof the initial crystallites be uniform and that the initial poly-dispersity in size be relatively small.


4/10

Synthesis of CdE Semiconductor Na nocrystallites J . Am. Chem. SOC.,Vol. 115, No . 19, 1993 8709

300 400 500 600 700Wavelength (nm)

Figure 2. Room temperature optical absorption spectra of -20-30 Adiameter CdS, CdSe, and CdTe crystallites.Surface Exchange. Preliminary studies of surfac e exchangehave been used to tailor th e com patibility of crystallites with avariety of solvents. Exposu re of purified crystallites to a largeexcess of pyridine begins th e exchange of the surface cap. Additionof solvents compatible with TOPO and T OP but incompatiblewith the new c ap results in the flocculation of the crystallites andthe removal of displaced TOP and TOPO species. Th e repeateddispersion in pyridine a nd isolation with alkanes drives the surfac eexchange with mass action. Crystallites capped with pyridinear e dispersible in polar solvents and arom atics but not aliphatics.Surfac e exchange results in a slight decrease in a verage crystallitesize and a small broadening of the size distribution, probably dueto the loss of species containing Cd and Se. Sh arp optical featurescan be recovered by size-selective precipitation from pyridinewith the titration of hexane. Crystallites can be stabilized in avariety of solvents with a range of functionalized caps. The re isobvious poten tial for customiz ing photochemical activity of theserobust chromophores through manipulation of their immediatechemical environment.Optical Properties. The absorption spectra of -20-30 Adiameter C dS, CdSe, a nd Cd Te nanocrystallite samples are shownin Figure 2. All three clearly show the effect of quantumconfinement. CdS, C dSe, and CdT e absorptions are shifteddramatically from their 512, 716, and 827 nm bulk band gaps,respectively. Th e CdS e spectrum shows three clearly resolvedtransitions while the Cd S and C dTe sam ples show less structure.W e do not believe the difference in quality of the optical spectrareflects any fundamental material limitations but rather theamount of effort spent in optimizing the growth conditions foreach material.Figure 3 shows the evolution of th e optical spectru m with sizein a series of room temperature absorption spectra for CdSecrystallites ranging from - 2 to 115 A in diameter. The seriesspans a ran ge of sizes from nearly molecular species containingfewer than 100 atoms to fragm ents of the bulk lattice containingmore than 30 000 atoms. Figure 4 compares the experimentallyobserved HOMO UM O gap as a functio n of particle size withthe prediction of the simple effective m ass approxim ation withthe Coulomb interaction treated in first-order perturbation.ld Allparticle sizes were determined by TE M and confirmed by X-rayline-shape analysis. Although experime nt and theory agreereasonably well at large sizes, the simple theory diverges fromthe expe rimental values for sm all sizes as expected fro m the non-parabolicity of the bands at higher wave vectors and the finitepotential barrier at the surface of the particles. Lippens and

ioWavelength (nm)Figure 3. Room temperature optical absorption spectra of CdSenanocrystallitesdispersed in hexane and ranging in size from - 2 to 115A.

6.05.55.04.5

5 4.0Pg 3.5ul

3.02.52.01.5 20 40 60 80 100 120

Diameter A)Figure 4. HOMO LUMO ransition energy of CdSe crystallites as afunction of size (diamonds) compared with the prediction of the effectivemass approximation (solid line).Lann ooL4 ave shown that tight binding calculations can yieldbetter agreement for these smaller sizes.Figure 5 shows the room temperature photoluminescencespectrum of a sample of 35 A diameter CdSe crystall ites andcompares it with the absorption spectrum. Th e luminescencequan tum yield for this samp le is -9.6 relative to R hodamine640 at room tem perature . Th e line width in emission is equal tothat in absorption, with the peak of the emission shifted 4 nm t othe red of the absorption maximum. This shift is the result ofa com bination of relaxation into shallow tra p states an d the sizedistribution.2e No deep trap emission w as detected.

(14) (a) Lippens, P. E.; Lannoo, M. Phys.Reu. B 1989,39, 10935. (b)Lippens, P. E.; Lannoo, M. Muter. Sci. Eng. B 1991, 9, 485.


5/10

8710 .I.m. Chem Soc.. Vol 115. No 9.1993 Murray et al.

400 500 Bw 700Wavelength (nm)

lm 5. Typical mom trmpcraturc hnd edge lumincsccncc andabsorption spxtra for 35 A diameter CdSc crystallites. No deep traplumincsssnce is detstcd.The sha rp absorption eatures suggest highly monodispersesamples. Hig h quan tum yields and narrow emission line widthsindicate growth of crystallites with few electronic defect sites.Th e sharp luminescence is a dram atic example of the efficiencyofth ecap ping grou pin electronically passivating thecrystalli tes.Th e capping groups also protect the individual crystallites fromchemical degrad ation yieldingrohust systems. Samp lesstored inthe original growth solution still show strong, sharp emissionafter storage for more than a year.These room temperature optical experiments, carried out oncommon lahoratoryequipment, dem onstrate thebenefits ofhigh-quality samples and point to the potential of more sophisticatedoptical studies on these samples.Transmission Electron Mieroseopy. Transmission electron

microscopy allows imaging of individual crystallites and thedevelopment of a statistical description of the size and shape ofthe particles in a sample. High mag nification imagingwith latticecontrast allows the determination of individual crystallite mor-phology.IsImaging at 3yu x IU times magnirication wm moocratecrystallite coverage allows careful size m easurements of 30to 50individual n anocrystallites on a single image and shows that theparticles are no t aggregated. Figure 6 shows a collection of slightlyprolate CdS e nanocrystallitesavexaging 35 A i in the directionof the ( 0 0 2 ) wr tz i t e ax i s and 30A f perpendicular to the(002) axis. Particles with the (002) axis perpendicular to thegrid ar e identified by a characteristic hexagonal pattern in atomimaging and a re nearly circular in cross ection. P articles orientedinotherdirectionsappearslightlyoblong. heanalysisofsamplesrangingfrom -2Oupto I15Aindicatesslightlyprolateprticleswithaspect ratioshetween 1.1 and 1.3.16 Samplesgenerallyhavestandard deviations


6/10

Synfhesis of Cd E Semiconductor Nonocrystallires

0 n mFiyw 8. The crystallite on the left ( - 1 10 A) shows a loss of latticecontrast in t he (101) pl an s pointing to the presen ceof4 stacking faultsalong the (W2) irection. The crystallite is prolate with an aspcct ratioof--1.3. Thccrystalliteontheright prsentsaprojectionperpendieularto the (002) direction and displays t h e hexagonal atom imagingcharacteristics of the wurtzite structure.of disorder in bulk 11-VI materials.a Chaderton et 01. also usechanges in lattice contrast to identify individual stacking faultsin a series of 10 0-nm 11-VI thin films.I8There is a broad distribution in the position and frequency ofstacking faults within each sample. Statistically more stackingfaultsareobse rved near thecenter ofcrystallites which may pointto an increased probability of faulting early in the growth. It isalso possible tha t greater contrast in the central region improvesthe detection of faults. Th e wavy lattice patterns of individualfaulted crystallites seen in TE M c an he qualitatively reproduccdincomputer generated atom images of crystallites by introducingstacking faults along the (002) axis.TEM allows not only the investigation of the morphology ofindivid ual crystallites bu t also the observation of ordering ofcrystallites intoseco ndary structur es. Twc-dim ensionalhexagonalclose packing heginstofo rm (F igu re9 )at relatively highcoveragelevels. Th e tailoring of mixed solvents in the fashion permittingselectiveprecipitation may allow more controlled growth of suchsecondary Structures.Selected area electron diffraction patterns co nfirm a predom-inantly wurtzite structure. Although a common technique,selected area diffraction is not emphasized in our structuralanalysis. Bright field images indicate that our crystallites arenot randomly oriented on he TE M grid but exhibit a tendencyto orient with their long axis parallel to the carbon surface. T henonstatisticaldistribution of crystal o rientationscan greatly impairthe interpretation of electron diffraction line shapes. Electronmicrodiffractionmay provide an appropriate alternative toselectedarea diffraction in these sy ~t em s. ~X-ray Diffraction. Unlike TEM, X-ray powder diffractionprobesa large numberofcrystallites thatarestatisticallyoriented.Samples are prepared as dense pellets free of any amorphousbinder topro videa strong signal with low background. Th e pelletsdisperse readily in a variety of solven ts with no pparent changein the absorption spectra.Samples of CdS, CdSe, and CdTe crystallites all exhibit apredominantly wurtzite crystal struc ture with the lattice spacing

(17) Awn. M ; rener. 1.S.Physics r?ndChrmirrryof 11-VI Compounds;18) Chaddcrton. L. T.; Fitrgerald. A. G.; Yoffe, A. D orure 1963.198,North Hollond Amsterdam, 1967;pp 144 a) and 132 b).=**_I,>. (19) Loretta.M . Elccrron beom aw ly i ofMorerio ls; Chapman andHall: London. 1984; pp 50-53.

1 Am . Chem.Soc.. Vol. I I S . No. 19. 1993 8711

mnmFi- 9. A near monolayer of 5 1 A diameter CdSe crystallitesshowingshort-range hexagonal close packing.

0 10 20 30 40 M 60 70 80 9028

0

Fi- 10. Powder X-ray diffraction spcctraof -35 A CdS. CdSc. andCdTecrystallits. The pasitions of bulk refledions far wnz i t e CdSarcindicated.of the bulk materials (Figure 10). Bulk crystals of CdS andCdS e commonly exhibit the wurtzite structure when prepared athigh temperature. The hexagonal modification of CdTe has beenobserved in thin films although to our knowledge it has not beenseen in bulk crystal^. ^Experimental X-ray powder diffraction spectra for CdSecrystallites ranging from - 2 o 115 A in diameter ar e displayedin Figure 11. Spe ctra show evidence of finite size broadenin g inall reflections? Excessive attenuation and broadening in the

(20)Guinicr.A.X-RoyDijrroclion:W.H.Freeman: SanFrancism. 1963:pp 122-149 (a) and 226237 b).


7/10

8712 J . Am. Chem. SOC., ol. 115, NO 9, 1993 Murray et al.Fits to experimental spectra employ a discrete form of theDebyeA

0 10 20 30 40 50 60213

Figure 11. Powder X-ray diffraction spectra of (a) 12, b) 18, (c) 20,(d) 37, (e) 42, f ) 83, and g) 115 A diameter C dSe nanocrystallitescompared with the bulk wurtzite peak positions (h).(102) and (103) reflections are c haracte ristic of stacking faultsalong the (002) axis.20b The diffraction pattern of the -12 Adiame ter species appears drama tically different from that of theother sizes. These sm all clusters possess too few atoms to de finea core crystal structu re, making the distinction between wurtziteand zincblende meaningless. Th e dram atic change in theirdiffrac tion eature s may indica te significant surface reconstructionor contributionsfrom thecapping g roups to thediffraction pattern.Proper structural analysis of these nearly molecular species willrequire the isolation of materials as single crystals as has beenachieved with Cd SZ 1StructuralSimulations. Modeling of nanocrystallite samplesas collectionsof m onodisperse spherical single crystals with idealsurface termination is appealing but may potentially be mis-leadin g. In this section we develop a working description ofaverage crystallite morphology using a com bination of expe ri-mental and computer simulated X-ray diffraction patterns andparameters which are consistentwith other independentstructuralprobes.

The combination of finite size and defect broadening resultsin a convolution of peaks in the X-ra y diffrac tion spec tra. Direc tobservation of peak positions and peak widths are thus un-reliable measures of lattice spacing and crystallite size. Butstructural information can be extracted by using TEM andEX AF S22 observations to guide in fi t t ing the experimentalX-ray diffraction spectra. Previous studies of the X-ray powderpatterns of CdSe nanocrystallites have demonstrated the sen-sitivity of spectra to the presenc e of plana r disorder and ther-mal effects.1 These studies are extended here over a rang e ofsizes. Th e importa nce of crystallite shape, the question oflattice contractions, and probable su rface disorder are investi-gated.

(21) Herron,N.;Calabrese, J. C.; Farneth, W. E.;Wang, Y . Science 1993,259, 1426. Our smallest CdS particles have optical spectra identical withthose in this reference. We believe the smallest CdSe particles in our seriesare the CdSe analogue to the CdS particles crystallized in this work.(22) (a)Marcus,M.A.;Flood, W.;Steigerwald,M.L.;Brus,L.E.;Bawendi,M . G . J .Phys. Chem.1991,95,1572. (b)Marcus,M.A.;Brus,L.E.;Murray,C. B.; Bawendi, M. G.; Prasad, A.; Alivisatos, A. P. Nanostruct. Muter. 1992,I 323.

Z(S) = zo j:*) in(2nrf i )2 * swhere Z(S) is the diffracted intensity, ZO s the incide nt intensity,AS is the scattering factor, and S is the scattering param eter[S= 2 sin(q)/X for X-rays of w avelength X diffracted throughangle q ] . Th e sum is over all discretized interatomic distancest k , and p(rk) is the num ber of times a given interatomic distancerk occurs. Since the number of discrete interatomic distances inan ordered structure grows much more slowly than the totalnumber of distances, the discrete form of the equation issignificantly more efficient for large crystallites.

Atomic coordinates are ob tained by system atically generatingatomic positions for a bulk crystalline lattice and retaining onlythose atoms falling within a defined ellipsoid. EX AF S studiesz2support the use of bulk bond lengths (2.63 A for CdSe). Thedimensions for the ellipsoid are taken from T EM measurem entsof average size and shape. Planar disorder along the (002) axisis reproduced by creating a collection of coordinate files with abroad distribution of stacking faults about the center of thecrystallites. Th e resulting spectra are statistically weighted andsumm ed to create a crystallite sample distribution with a definedaverage defect density. Thermal effects are simulated by theintroduction of a DebyeWaller factor.24 A constant Debye-Waller fac tor is used based on a me an-square displaceme nt of0.04A2 for each atom, consistent with EXA FS studies22whichfind bulk De bye Wa ller factors. The Debye-Waller factor doesnot chan ge peak positions and only affects peakintensities throughan exponential damping as 28 increases. The broad backgroundsin the ex perimental diffraction spectra are subtly influenced bythe combined effects of experimental geometry, variable ab-sorption of X-rays, contributions from incoherent scattering,therm al effects (Debye-Waller term), and scattering from thecapping groups. Rigorous correction of the background fromthese contributions requires a knowledge of sa mple propertiesnot presently available. A constant b ackground chosen to bestfit the relativeintensityof the (1 10)reflection for 37 Acrystallitesis added to all simulated spectra to crudely compensate for theexperimental contributions to the background. This backgroundis not important in our analysis which concentrates on thediffraction features.

Figure 12compares calculated spectra of 1000 atom sph ericalCdSe part icles (-37 A diame ter) with the experimental powderpattern of a similarly sized sample. Th e spectra of defect freezincblende (a) and w urtzite (b) crystallites do not reproduce theoverall shape of experimental spectrum d, especially in the regionbetween the 1 10) and (1 12) peaks. The introduction of a singlestacking fault near the ce nter of an otherwise wurtzite crystallitegreatly improves the fit (spectrum c) by broadening the (103)peak.The (110)spacing is present in both the wu rtzite and zincblendestructures and is thus unaffected by the presence of stackingfaults along the (002) axis. This spacing also has reasonableseparation from neighboring reflections. Crude estimates ofparticle size based on the Schere r analysis of peak widths shouldemploy the width of the (110)feature rathe r than the convolutionof the (loo), (002), and (101) reflections found in the firstdiffrac tion feature . This convolution is easily confused with the

(11 1 ) zincblende reflection, and using it to estimate p article sizewould underestimate the coherence length. Th e agreement inthe width and shape of the (1 10) feature between our simulatedspectra using crystallite dimensions based on TEM measurem ents(23) Hall, B. D.; Monot, R. Comput. Phys. 1991, 5, 414.(24) Vetelino, J. F.; Guar, S . P.; Mitra, S . S . Phys. Reu. B 1972,5,2360.


8/10

Synthesis of CdE Semiconductor N anocrystallites J . Am. Chem. Soc. , Vol. 115, No. 19, 1993 8113I I I

10 20 30 40 50 60 7020Figure 12. Simulated X-ray powder diffraction spectra for 35 A diameterspherical nanocrystallites: (a) pure zincblende, (b) pure wurtzite, (c)wurtz ite with one stacking fault. (d) Experimental powder spectrum of-35 A diameter crystallites.and o ur experimental sp ectra confirms crystallite size and indicatesthat structural defects must be coherent defects (e.g. stackingfaults).Figure 13 highlights the sensitivity of the diffraction spectrumto crystallite shape. Th e experimental spectrum of a sam ple ofCdSe nanocrystallites containing on average - 0 000 atoms(-80 8 diameter) is compared with simulations. Th e best fitsobtained assuming spherical particles (spectrum a ) and prolateparticles with an aspect ratioo f 1.3 in the (002) irection (spectrumb) ar e shown. Each simulation has an average defect density of3 stacking faults per crystallite to fit the region of the (1 03) peak.While the (100) feature is more intense than the (002) peak inbulk Cd Se and in th e simulated spectrum of spherical particles,the opposite is true fo r our crystallites and in the sim ulation ofprolate particles. The elongated particles have a larger numberof (002) planes, making tha t peak the dom inant reflection in thefirst diffraction feature. Th e greater coherence length in thatdirection results in the narr owin g of (002) reflection and a smalleroverlap with theneigh boring ( 100) and (101) reflections. Inclusionof th e prolate natu re of the particles, observed in T EM , is essentialfor reproducing the shape of the convolution of the (l oo ), (002),and (101 ) reflections. W e fix the aspect ratio at -1.3 for thesimulations of the other sizes to minimize the num ber of floatingparameters.Th e first broad diffraction featu re (28 - 5') app ears to shiftto higher sc attering angles with decreasing particle size (Figure11). An isotropic lattice contraction of a few percent due tosurface tension has been proposed as a possible explanation forsimilar observations in C dS nanocrystallites.25 EX AF S dataZ 2however yield averag e bond lengths which ar e essentially identicalwith those in the bulk even for the smallest crystallites. In fact,even the small - 28 Cd S clusters recently crystallized by He rron

(25) (a) Wang, YAHerron, N . Phys. Rev.B 1990,42,7253. b)Goldstein,A. N.; Echer, C. M ; livisatos, A. P. Science 1992, 256, 1425. (c) Colvin,V. .; Goldstein,A. N.; Alivisatos, A. P. J . A m . Chem. SOC. 992,114,5221.

002

002

15 20 25 30 35 40 45 50 5528

Figure 13. Experimental X-ray powder diffraction spectrum of -80 Adiameter - 10000 atoms) crystallites (dotted line) compared withcomputer simulations solid line)of (a) sphericaland (b) prolate particles[aspect ratio of 1.3 along the (002) irection]. Three stacking faults onaverage and bulk lattice constants are used.

20 25 30 35 40 45 50 5528

I20 25 30 35 40 45 50 5528

Figure 14. Experimental X-ray powder diffraction spectrum of -20 A(-275 atoms) diameter crystallites (dotted line) compared with a seriesof computer simulations of wurtzite crystallites containing 275 a tomwith one stacking fault per crystal lite in the (002) irection (solid lines).(a) Spherical particles with bulk lattice parameters,(b) spherical particleswith a 2% isotropic lattice contraction, (c) prolate particles [aspect ratioof 1.3 along the (002)axis] with bulk lattice parameter, (d) moderatesurface disorder added to the prolate particles of part c.et al.21 only show a bond contrac tion of -0.5% compared to thebulk. This apparen t contradiction can be resolved in our samplesby properly taking crystallite shape into account. Figure 14acompares the exp erimental spectrum of small CdS e crystallitescontaining -275 atoms (-20 A diameter) to the simulatedspectrum of spherical particles with a bulk lattice parameter,showing a clear shift of the first experimental peak to higherangles. Figure 14b compares the same experimental spectrumwith the sim ulated spectrum of spherical particles with a lattice


9/10

8714 J. Am . Chem. SOC.,Vol 115, No. 19, 1993param eter contracted by 2%. A constant defect density of onestacking fault/crystallite is used in both simulations to fit thevalley between the (1 10) and 112) peaks. The 2% contractionis designed to reproduce the position and general shape of thefirst diffraction feature b ut results in a poor fit at higher angles.Observation of a shift in the first peak is not enough to prove alattice contraction. Th e higher peaks must also shift. Figure14c shows the effect of putting a n aspect ratio on the crystallites.Theexperimental pa ttern iscompared with thesimulated spectrumof prolate particles having an aspect ratio of - .3 in the (002)direction, a bulk lattice parameter, and one stacking fault percrystallite. The introduction of crystallite shape uniformlyimproves the fit without requ iring a major lattice contraction.The prolate shape results in th e promine nceofthe (002)reflection,skewing the first diffraction fe ature to higher scattering angles.This effect is especially impo rtant for sm aller crysatllites whichhave few planes and where a single extra plane has a tremendouseffect on the intensity and w idth of the diffraction feature forthat direction. Thus, for the sam e aspect ratio, small crystallitesexhibit a larger appare nt shift tha n larger crystallites. Fits withprolate crystallites reproduce the position of the peaks a nd thegeneral shape of the spe ctrum but show a lack of intensity in abroad region between 28 - 27' and 33O compared to theexperimental da ta.

TEM observations suggest poorly resolved facets and thepresence of a potentially reconstructured surface layer limitingthe accuracy of measurements to no better than one atom ic plane.Th e presence of surface disorder is simulated in our studies bythe relaxation of all atoms lacking their full coordination shell.Th e poorly defined crystal surfaces make the choice of a specificreconstruction difficult. Studies by Duke and Wa ng on clean11-VI surface s in high vacuum h ave establishe d gener al mech-anism s for reconstruction on cleavage faces.26 A general tendencyfor a bond length conserving rotation of surfa ce 11-VI pairs resultsin the cation relaxing into the surfa ce while the anion is pushedout of the plane. W e retain the character of these relaxations inour simulations by radially trans lating all surface cations 0.40 Atoward the cylindrical axis of the crystallite. The surface anionsare translated outward byO.2OA. Th e magnitudeof therelaxationis comparable to that of Duke a nd W ang26 and it is chosen toprovide the best fit to the experimental data . The average bondlength is kept to within f0.0 05 of the bulk value for consistencywith EXAFS experiments which show negligible bond lengthcontractions.22 Figure 14d shows that a greem ent with experi-me ntal spec tra can be improved by the introduction of such surfacedisorder. Th e choice of the relaxation is not unique and is notan a ttem pt to define a specific surface structure, rather it is toshow that su rface disorder is compatible with the experimentalX-ray data . We fix the amou nt of surface disorder to that usedfor the 2 0 4 art icles above since we do not expect i t to be stronglysize depe ndent. Effec ts of surface disorder become imperceptibleas the crystallite size increases (>SO A).

Figure 15 shows the experimental spectrum of a sam ple of 37A diameter crystallites - 1000 atoms) compared with a sim-ulation for crysta llites with a defe ct density of 1.3 stacking faultsper crystallite, an aspect ratio of - .3along the (002) direction,and a relaxed surface. These particles ar e small enough thatsurface contributions cannot be completely ignored. All threedeviations from ideality are required to properly fit the fullspectrum. Each deviation affects the spectrum independentlyand in different regions. Th e stacking faults mostly affect thearea between 28 - 0 and SOo, the prolate sha pe affects thepeak position of the first diffraction feature a t 26 - 5O, and thesurface disorder affects the shape of th e first feature.

Murray et al.

(26) (a) Duke, C. B.; Patton, A.; Wang, Y. R.; Stiles, K.; Kahn, A. Surf .Sci. 1988 , 197 , l l . b) Wang Y.R.;Duke C.B.Phys.Reu.B1988 37 6417.(c) H orsky, T. N.; Brandes, G. R.; Canter, K. F.; Duke, C. B.; Paton, A.;Lessor, D. L.;Kahn, A.; Horng, S. .; Stevens, K.; Stiles, K.; Mills, A. P.,Jr. Phys. Rev. E 1992, 46, 7011.

15 20 25 30 35 40 45 50 55 6028

Figure 15. Experimen tal X-ray powder diffraction spectrum of -37 A(-lo00 atoms) diameter crystallites (dotted line) compared with asimulated spectrum (solid line) for wurtzitecrystallites containing O00atoms with one stacking fault per crystallite in the (002) direction, anaspect ratio of 1.3 along the 002) axis, and a small amount of surfacedisorder.

In sum mary, powder X -ray diffraction patterns of a series ofCdSe samples can be fit with computer simulations usingparameters consistent with TEM and EXAFS investigations.These simulations assume a bulk c rystal structure and rely on asmall number of parameters (seven) to develop a workingdescription of avera ge crystallite morphology. These param etersar e the lattice constant, the amoun t of thermal disorder (Debye-Waller facto r), crystallite size, aspect ratio, stacking fault defectdensity, the am ount of surface'disorder, anda constant backgroundcorrection. As noted above, eac h para me ter affects the powderpattern independently and in a different region, giving a robustfit. Of these seven parameters, two are fixed from TE M andEX AF S data: The lattice constant is fixed at the bulk value(EXA FS) and the crystallite size if taken from T E M measure-ments. Of the rema ining five parameters, four (aspect ratio,surf ace disorder, background corre ction, Debye-Waller facto r)are fit for one crystallite size and then fixed at tha t value for thewhole series of sizes. As noted, these four parameters are alsoconsistent with independent structural probes. Th e remainingparam eter, the average stacking fault defect density, is allowedto float from size to size. Its value is however relatively constantover the entire series of sizes.The com bination of X-ra y studies and TE M imaging yields adescription of average CdSe nanocrystallite structure. Stric tclassification of the struc ture as purely wur tzite or zincblende ispotentially misleading. T E M images of crysta llites grown slowlya t high tem perature provide dimensions of slightly prolate particlesand indicate the presenceof stack ingfaults in th e (002) direction.Diffraction patterns a re consistent with a predominantly wurtzitestruc ture avera ging one stacking fault defec t every 6-7 planes inthe elongated (002) direction. Th e presence of planar disordermay be important in understanding the influence of structure onth e electronic properties an d phase s tabilities of nanocrystallites.The prolate shape results in a greater coherence length in the(002) direction and provides a potential explanation for theapparent shift in the first diffraction peak without requiringsignificant lattice contractions or phase transitions. A smallamount of surface disorder, consistent with TEM nd EXAFSobservations, improves agreement between simulated and ex-perimental diffraction spectra but does not uniquely define asurface structure.


10/10

Synthesis of CdE Semiconductor N anocrystallitesIV. Conclusion

A relatively simple route to the synthesis of macrosco picquan tities of nearly monodisperse 11-VI sem icondu ctor nano-crystallites is presented. The techn ique allows an e ntire seriesofsamples, ranging from - 2 o 115A in diameter, to be obtainedin a single reaction and in macrosco pic quantities. Th e particlesproduced have uniform size, shape, and su rface passivation andshow relatively sharp absorption and em ission feature s at roomtemperature.The combination of TEM imaging and comparisons ofexperimental and simulated X-ray diffraction spectra providesa self-consistent descr iption of crystallite structur e. The se studieshighlight the importance of shape and disorder in interpretingexperimental results. Average bulk bond lengths and a pre-dominantly hexagonal (wurtzite) crystal struc ture are found, evenin crystallites containing as few as 275 atoms (- 20 A diame ter).

J . A m . Chem. SOC.,Vol 115, No. 19, 1993 8715The synthetic rationale and systematic structural analysispresented here should be generalizable to the produc tion of avariety of new nanocrystalline materials.

Acknowledgment. We thank Mike Frangillo and BashirDabbousi for valuable assistance with the transmission electronmicroscopy, Peter Nelson for helping set up the computermodeling, and Vicki Colvin for useful discussions. C.B.M . andD.J.N. gratefully acknowledge fellowships from NSERC andNS F, respectively. This research was funded in part by the MI TCenter for Materials Science and Engineering (NSF-D MR-9 0-22933), N S F (DM R-91-57491, CHE-89-14953, and ECS-91-18907), he donors of the Petroleum Research Fund, adm inisteredby the Ame rican Chem ical Society (ACS-PR F-24398-G6), andthe Lucille and David Packard Foundation.

murray, norris y bawendi

Documents