japplphys 60 alp

8/3/2019 JApplPhys 60 Alp

1/18

Atomic layer epitaxyColin H. L. Goodman and Markus V. Pessa

Citation: J. Appl. Phys. 60, R65 (1986); doi: 10.1063/1.337344View online: http://dx.doi.org/10.1063/1.337344

View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v60/i3

Published by the American Institute of Physics.

Related Articles

Influence of the interface on growth rates in AlN/GaN short period superlattices via metal organic vapor phaseepitaxyAppl. Phys. Lett. 99, 201903 (2011)

Extremely high absolute internal quantum efficiency of photoluminescence in co-doped GaN:Zn,SiAppl. Phys. Lett. 99, 171110 (2011)

Effect of graded-temperature arsenic prelayer on quality of GaAs on Ge/Si substrates by metalorganic vaporphase epitaxyAppl. Phys. Lett. 99, 171908 (2011)

GaN/Al0.5Ga0.5N (11-22) semipolar nanostructures: A way to get high luminescence efficiency in the nearultraviolet rangeJ. Appl. Phys. 110, 084318 (2011)

Surface transition induced island formation on thin strained InGaN layers on GaN (0001) in metal-organic vapourphase epitaxyJ. Appl. Phys. 110, 073527 (2011)

Additional information on J. Appl. Phys.

Journal Homepage: http://jap.aip.org/

Journal Information: http://jap.aip.org/about/about_the_journal

Top downloads: http://jap.aip.org/features/most_downloaded

Information for Authors: http://jap.aip.org/authors

Downloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
http://jap.aip.org/search?sortby=newestdate&q=&searchzone=2&searchtype=searchin&faceted=faceted&key=AIP_ALL&possible1=Colin%20H.%20L.%20Goodman&possible1zone=author&alias=&displayid=AIP&ver=pdfcovhttp://jap.aip.org/search?sortby=newestdate&q=&searchzone=2&searchtype=searchin&faceted=faceted&key=AIP_ALL&possible1=Markus%20V.%20Pessa&possible1zone=author&alias=&displayid=AIP&ver=pdfcovhttp://jap.aip.org/?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.337344?ver=pdfcovhttp://jap.aip.org/resource/1/JAPIAU/v60/i3?ver=pdfcovhttp://www.aip.org/?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3658734?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3655678?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3656737?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3654053?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3647782?ver=pdfcovhttp://jap.aip.org/?ver=pdfcovhttp://jap.aip.org/about/about_the_journal?ver=pdfcovhttp://jap.aip.org/features/most_downloaded?ver=pdfcovhttp://jap.aip.org/authors?ver=pdfcovhttp://jap.aip.org/authors?ver=pdfcovhttp://jap.aip.org/features/most_downloaded?ver=pdfcovhttp://jap.aip.org/about/about_the_journal?ver=pdfcovhttp://jap.aip.org/?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3647782?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3654053?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3656737?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3655678?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.3658734?ver=pdfcovhttp://www.aip.org/?ver=pdfcovhttp://jap.aip.org/resource/1/JAPIAU/v60/i3?ver=pdfcovhttp://link.aip.org/link/doi/10.1063/1.337344?ver=pdfcovhttp://jap.aip.org/?ver=pdfcovhttp://jap.aip.org/search?sortby=newestdate&q=&searchzone=2&searchtype=searchin&faceted=faceted&key=AIP_ALL&possible1=Markus%20V.%20Pessa&possible1zone=author&alias=&displayid=AIP&ver=pdfcovhttp://jap.aip.org/search?sortby=newestdate&q=&searchzone=2&searchtype=searchin&faceted=faceted&key=AIP_ALL&possible1=Colin%20H.%20L.%20Goodman&possible1zone=author&alias=&displayid=AIP&ver=pdfcovhttp://aipadvances.aip.org/?ver=pdfcovhttp://jap.aip.org/?ver=pdfcov


2/18

Atomic layer epitaxyColin H. L. GoodmanStandard Telecommunication Laboratories, Harlow. Essex. EnglandMarkus V. PessaDepartment 0/Physics, Tampere University 0/ Technology. SF-33JOJ Tampere. Finland(Received 4 November 1985; accepted for publication 31 March 1986)Atomic layer epitaxy (ALE) is no t so much a new technique for the preparation of thin films as anovel modification to existing methods of vapor-phase epitaxy, whether physical [e.g.,evaporation, at one limit molecular-beam epitaxy (MBE)] or chemical [e.g., chloride epitaxy ormetalorganic chemical vapor deposition (MOCVD) ]. It is a self-regulatory process which, in itssimplest form, produces one complete molecular layerof a compound per operational cycle, witha greater thickness being obtained by repeated cycling. There is no growth rate in ALE as in othercrystal growth processes. So far AL E has been applied to rather few materials, but, in principle, itcould have a qui te general application. It has been used to prepare single-crystal overlayers ofCdTe, (Cd,Mn)Te, GaAs and AlAs, a number of polycrystalline films and highly efficientelectroluminescent thin-film displays based on ZnS:Mn. It could also offer particular advantagesfor the preparation of ultrathin films ofprecisely controlled th ickness in the nanometer range andthus may have a special value for growing low-dimensional structures.

TABLE OF CONTENTSI. IntroductionII. Film growth by atomic layer epitaxyIII. Theoretical aspects

A. Growth modelsB. Kinetic effects

IV. Growth by atomic layer epitaxy-experimentA. Monocrys talline deposits

1. Cadmium telluride2. Cadmium manganese telluride3. Gallium arsenide

B. PolycrystaHine and amorphous deposits1. Zinc sulfide2. Zinc teHuride3. Oxides

V. The extension of ALE to other materialsVI. Conclusions

J. INTRODUCTIONThin films are of the greatest technological importance,

and range widely in application: the active layers in semiconductor devices as well as the dielectrics in them or in electrolytic capacitors, the transparent conductors in liquid-crystaldisplays or on aircraf t windshields, the luminescent and protective layers in electroluminescent (EL) thin-film displays,etc. More recently ultrathin films of uniform and preciselycontrolled thickness in the range 1-10 nm have come to be ofthe greatest importance both scientifically in the branch ofsolid-state physics termed "low-dimensional," as well astechnologically in integrated circui ts and optoelectronic devices.

In this general context little attention was initially paidto a patent of Finnish origin taken out in 1977 1 which described a novel mode ofevaporation deposition, atomic layerepitaxy (ALE), for preparing thin films of zinc sulfide. This

originated from the Lohja Corporation bu t was probably tounorthodox and seemingly restrict ed in application to makmuch impact. By 1980, however, it was being used to makremarkably good large-area thin-film EL displays based oZnS:Mn (Refs. 2 and 3) , and in 1982 still better displayfrom Lohja won, against strong U.S. and Japan competitionthe prize of the Society ofInformation Display for the greatest advance in the display field during the preceeding yearALE began to attract considerable attention as a method foproducing thin films of high quality. More recently still, witthe epitaxial film work of Pessa's group at Tampere University of Technology, Finland, on n-VI compounds, and oNishizawa's group at Tohoku University, Japan, on galliumarsenide, as well as works of Usui et al. at Nippon ElectriCorporation and Bedair et al. at North Carolina State Unversity, on GaAs, the promise of AL E for growing low-dmensional structures has generated much interest.II. FILM GROWTH BY ATON'llC LAYER EPITAXY

A particularly simple way to describe the AL E approach is to say that it makes use of the difference betweechemical absorption and physical adsorption. When the firslayer of atoms or molecules of a reactive species reaches solid surface there is usually a strong interaction (chemisorption); subsequent layers tend to interact much lesstrongly (physisorption). I f the initial substrate surface iheated sufficiently one can achieve a condition such thaonly the chemisorbed layer remains attached. In the earliesand perhaps simplest example of ALE, viz., the growth oZnS by evaporation in a vacuum, Zn vapor was allowed timpinge on the heated glass. Evaporation was then stoppeand ifphysisorbed Zn were present because of the impinginflux, it would reevaporate. Th e process was then repeatewith sulfur, the first layer of which would chemisorb on thinitial Zn layer; any subsequent physisorbed sulfur woulgradually come away from the heated substrate when th

R65 J. Appl. Phys. 60 (3),1 August 1986 0021-8979/86/150A65-17$02.40 @ 1986 American Institute of Physics A6Downloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions


3/18

0)substrate

BA M0\,. 0\0) b)

b)su bstrateII, ,I e\

c) d)subs trate I400U

+e:::>Ipu I se 1

- - - ~ . . . - . - . , . ifK;) G b ~ 1 .bJ I I I I Ipulse 2 I pulse 1

IFIG. L Film growth by ALE. Upper panel: growth from elemental sourcematerials [variant (i ) ). Lower panel: growth from compound source materials via surface exchange reactions [variant (ii)].

sulfur flux was cut off, leaving one (double) layer of ZnS.This complete cycle could be repeated indefinitely, the number oflayers grown being determined solely by the number ofcycles [see Fig. 1(a) ]. This cyclic vacuum evaporation wasdescribed in the original Lohja patent. 1

It has been shown by straightforward measurements4that chemical vapor deposition (CVD) of ZnS from ZnC1 2and H 2S could also be operated in an ALE mode, the cyclebeing somewhat more complex [Fig. t (b) 1 n that a chemisorbed ZnCl2 monolayer loses its chlorine to the hydrogen ofthe later-arriving H 2S molecules, forming HCI, and againresulting in a (double) layer of ZnS. The importance of thisCVD variant is that it enables ALE to be applied to compounds with one or more involatile constituents for whichthe evaporative approach would clearly no t be viable.One can thus arrive at a definition of ALE: it is based onchemical reactions at the solid surface of a substrate, towhich the reactants are transported alternately as pulses ofneutral. molecules or atoms, either as chopped beams in highvacuum, or as switched streams of vapor possibly on an inertcarrier gas. The incident pulse reacts directly and chemicallyonly with the outermost atomic layer of the substrate. Thefilm therefore grows stepwise-a single monolayer perpulse-provided that at least one complete monolayer coverage of a constituent element, or of a chemical compoundcontaining it, is formed before the next pulse is allowed toreact with the surface. Given that , any excess incident molecules or atoms impinging on the film do not stick i f the substrate temperature Tgt is properly chosen, and one thereforeobtains precisely a monolayer coverage in each cycle.

The formation of a "layer per cycle" is the specific feature that conceptually distinguishes the ALE mode fromother modes of vapor phase deposition; the l:atter all give agrowth rate, ALE gives a growth per cycle.R66 J. Appl. Phys., Vol. 50, No.3, 1 August 1986

Some confusion appears to have been generated becausethe ALE-grown materials actually used by Lohja in theirdisplays were polycrystalline or amorphous. Also it was noteasy from what was published to deduce experimental conditions. It was only later that the first controJJ.ed experimentsin a laboratory environment4s were carried out at the Tampere University of Technology in 1979-80, and providedconfirmation of the general correctness of the ideas developed by Suntola at Lohja. These were followed by furtherexperiments involving deposition onto single-crystal substrates, discussed later in Sec. IV. These experiments showedthat the ideaJized ALE picture, of one complete double layerbeing grown per cycle, could be an oversimplification. Asdiscussed subsequently, incomplete layer growth occurs inmany cases.

To sum up the present situation: the ALE mode of deposition has so far only been applied. to a rather small numberof materials, listed in detail in Sec. IV. These materials are ofconsiderable scientific and technological interest. It seemshighly probable, though yet unproved, that a very muchwider range of materials could be handled. One particularlyremarkable feature claimed for the ALE approach is that itautomatically gives an absolute control of deposit thicknessin terms of the number of cycles employed. However, although the general conditions under which ALE would bepossible seem reasonable clear, each material will have itsown specific requirements regarding the operational conditions under which deposits of high perfection and puritycould be obtained-while in some cases ALE might tum outnot to be applicable.

iii. THEORETICAL ASPECTSA. Growth models

As already indicated ALE in its basic form operates bygrowing complete atomic layers on top of each other, therebeing two basic variants based on (i ) evaporative deposition[with as one extreme, molecular-beam epitaxy (MBE) ], relying on heated elemental source materials, or (H ) chemical.vapor deposition (CVD), relying on sequential surface exchange reactions between compound reactants.

ALE, therefore, is not so much an entirely new methodof crystal growth as a special mode of these well.-establishedgrowth techniques.Both ALE variants, illustrated in Fig. t., assume that atthe substrate temperature the vapor pressures of the sourcematerials as solid phases are much larger, perhaps by severalorders of magnitude, th an those of he compound films eventually fonned. from them. This requirement is usually met inthe case of II-VI materials and can hold reasonably well inmany other cases.

The precaution must be taken to turn off the sourcebeam for a sufficiently long time (o f the order of 1 s) aftereach deposition pulse. In this way the surface is allowed toapproach thermodynamic equilibrium at the end of each ,re-action step-a growth condition that is not usually met inother techniques of deposition by evaporation. It is thenclear that the fluid dynamics are entirely eliminated from thegrowth problem. We should also note that a low net growth

C. H. L. Goodman and M. V. Pessa R66Downloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions


4/18

rate, a characteristic of ALE, can be partly compensated bydesigning reactors properly to allow for depositing severallarge-area samples simultaneously (such a reactor is illustrated in Fig. 4; see Sec. IV).In ALE variant (i), element A and element B are alternately deposited onto the substrate at temperature Tg r ' Thenumber of monolayers produced by one pulse can be calculated from

Here OJ is the atomic density for one monolayer coverage(0 = 1), Ii the number of atoms per second impinging onthe substrate of unit area, Ie the number of atoms per secondreevaporated from the substrate during deposition, and tjthe duration of the pulse. As () reaches unity the stickingcoefficient abruptly changes from unity to zero; i.e., Ie becomes equal to Ii ' Only a monolayer of element A held bystrong bonds is left behind after reevaporation of thoseatoms (o r molecules) that are present in excess. Next, thesurface is subjected to a pulse of B. Once all possible A-Bbonds have formed, no further B will stick. The process isrepeated with alternate pulses of A and B until the desiredfilm thickness is achieved, one monolayer at a time.Such behavior can be termed the ideal ALE mode.6 Experiments, however, indicate that this model is oversimplified because it does not take into account the fact that perfectlayer-by-Iayer growth does not always take place, even if anoptimum value of Tgr is chosen. A detailed study of growthofCdTeepilayerson CdTe (111) substrates has shown7 thatthe sticking coefficients of Cd and Te are less than unity atTgr The persistent coverages left behind after reevaporationare measured to be only 35% of a monolayer for Cd on CdTe(lIDB and 72% for Te on CdTe(111)A. Such coverageshave structural implications-see the discussion in Sec. II IB. A generalized model has been suggested7 for growth ofanABcompoundby AL.E variant 0) which involves (a) theexistence of weakly bound states of both reactants in thevicinity ofthe interface and (b) partial reevaporation of thefirst elemental monolayer deposited. According to this model, the timing and dose of the sequential pulses of elementalA and B should be carefully related to each other. This, ofcourse, represents a considerable departure from the idealALE model but does retain the basic growth per cycle that isthe distinctive feature of ALE.In the case of ALE variant (ii), sequential surface ex-change reactions are used. to grow a film of a compound AD[see Fig. 1(b) ]. During the first phase of a deposition cycleincident molecules AB are adsorbed by the surface layer. ABremains on the surface until a pulse of CD molecules arrives.At a properly chosen Tgr the surface reaction AB + CD-BC + AD then occurs. This releases Band C (usually as acompound BC) and results in the formation of a layer ofAD. As with ALE variant (i), the growth proceeds stepwise,and is independent of the size of a reactant pulse, providedthat it exceeds the minimum amount o f material required toform the appropriate adsorbed monolayer. I t nonethelessappears probable that, as noted for variant (i), the timingA67 J. Appl. Phys., Vol. 60, No.3, 1 August 1986

and dose of sequential pulses would need to be related teach other.Comprehensive theoretical quantum chemical studieof he growth of a ZnS surface by ALE variant (ii) using thfirst of these reactions have been carried out by Pakkanenal.8 These studies showed that the growth process, whicinvolves several reaction steps, again is not so simple as thideal ALE reaction scheme shown in Fig. lb. The calculations were based on ab initio Hartree-Fock theory for atomand molecules. They showed that the formation of a complete ZnS overlayer does not occur in a single ALE cycle.The calculations suggest that two different growtmechanisms could occur involving ALE-type cyclic reaction with molecules of ZnClz and HzS. These mechanismmight operate simultaneously at different sites of the reZnS surface. In the first of these mechanisms surface chainare formed, the first step in this chain process being the chmisorption of ZnClz which produces a chlorine-bridgeZnCl2 overlayer with a maximum surface coverage of 0.5Subsequent HzS adsorbates react with the ZnClz chains releasing HCI and creating ZnS chains with the S atoms at thtop. ZnCl2 molecules of the next pulse bind to the S atomand again form a chlorine-bridged ZnClz overlayer with surface coverage of0.5. The second H 2S pulse completes thnew ZnS layer. This complete layer is identical to the bulstructure. On this basis, the maximum growth rate is onnew layer per two cycles of ZnCl2 + H 2S. In the seconmechanism, an alternative, somewhat similar set of reations was proposed for the formation of ZnC12-ring ovelayers. The initial surface coverage would now be 0.33 anthe maximum growth rate would therefore become one Znlayer per three cycles of ZnC1 2 + H2S.Deviations from the simple model of ALE growth dicussed so far have aU indicated a reduction of the thicknegrown per cycle. However, some recently published work9on GaAs grown by ALE variant (ii), and discussed in Setion IV A 3, indicates that growth can proceed in a complelayer-by-layer fashion; in some circumstances there can evebe an increase of the thickness per cycle. I IB. Kinetic effects

Several factors must be satisfied for ALE growth to bpossible. ALE implies the need for (a) a flat perfect substrate surface and (b) equally strong adsorption of the twvapor species taking part in the cycle, and equally easy desorption of any reaction products in the case of ALE varia(ii). It does not as yet appear possible to calculate accurateto what extent these requirements will be met by a specifmaterial, simply because not enough is known about the efects due to surface adsorption of impurities (including adsorption of excess of constituent elements). There are, however, some useful guidelines.The flatness/perfection of a low-energy (close-packedsurface of a crystalline material depends on the energy required to form defects in it. Some insights into this can bobtained from Jackson's model l2 which demonstrates dose relationship between surface roughness during growtand entropy for transformation (i.e., volatilization in thcase of vapor growth, fusion for growth from the liquid). A

C. H. L. Goodman and M. V. Pessa ADownloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http:// jap.aip.org/about/rights_and_permissions


5/18

L f ' 1 ~ 4Za::UJzUJUJLLUJ>


6/18

directly reflects the surface kinetics that govern the adsorption and incorporation of Ga an d As into the growing crystal.

Considerable effort has been expended in trying to understand the detailed adsorption kinetics for GaAs surfaces,see, e.g., Refs. 16-19. However, although qualitative agreement with experiment has been obtained, as for example inRef. 18, which deals with the effects of adsorption on thevapor growth of GaAs, it should be noted that even for sucha well-characterized material as GaAs various assumptionshad to be made, while some of he relevant parameters had tobe plausibly estimated. It thus seems unlikely that it will bepossible in the near future to predict the orientation-dependent behavior of the vapor growth of any less well-understood materia1. At least for the present time. therefore.whether or not a particular material might be grown by ALEmust be judged semiempirically.IV.GROWTH BY ATOMIC LAYER EPITAXY-EXPERIMENT

The cycle of alternate pulses in ALE may be repeatedevery 1-3 s, which yields a net growth rate of about 0.3-0.6j.lmlh. Considerably longer cycle times, and hence slowergrowth rate, can also be used. but any significant reductionin cycle time below 1 s is unlikely, due to limited pump capacity. As already mentioned in Sec. III. the small net deposition rate may be compensated by simultaneously growingmany films. An ALE reactor capable of handling a batch ofsubstrates is illustrated in Fig. 4. Here the separation of reaction steps is accomplished by a gas flow over immobile substrates. The reactants are alternately injected into thegrowth chamber. and. following each reaction. are purgedaway by a carr ier gas flow.

A. Monocrystalline depositsThe first detailed investigation of atomic layer epitaxyon single-crystal substrates was carried out at the TampereUniversity of Technology.20 This employed an ultrahighvacuum apparatus of the kind used for MBE. In order tomaintain the vacuum of 10 -7 Pa during growth limitedpump capacity made it necessary to use a relatively longcycle time. The final result was a net growth rate of 0.1-0.3j.lmlh. Later work6.7.21.22 employed a new and improvedMBEIALE system, illustrated schematically in Fig. 5. andenabled faster growth to be achieved. The growth chamberincorporates three Knudsen-type effusion furnaces manufactured by Kryovak Ltd. for operation in the ALE mode (i )(and in the usual MBE mode) and two gas inlets for applying ALE variant (ii). This is built onto a VG ADES-400electron spectrometer capable of x-ray photoelectron

moin gas flo'ol

'-- . - - - - - y - ' ~ _ .. -'reoctants

heating elements substrates

pump

FIG. 4. Schematic diagram of a gas-flowtype ALE reactor.

R69 J. Appl. Phys., Vol. 60, No.3, 1 August 1986

~ P S UPSAES

~ ; ; ; ; ; ; ; ~ _ _ ...:_= hOrIZontalmanipulator

FIG. 5. Layout of an ultrahigh vacuum reactor chamber capable of ALEand MBE growth by evaporation from Knudsentype effusion cells or bsurface exchange reactions of gases from vapor sources, built on a multtechnique electron spectrometer for in situ characterization of films.

(XPS). Auger electron (AES), angle-resolved UV photoemission (UPS), and low-energy electron diffraction(LEED) measurements.

1. Cadmium tellur ideThe first material to be investigated was CdTe. 2o AL E

variant (i ) was used because ofthe rather high vapor pressures of elemental Cd and Te at elevated temperatures wherthe dissociative evaporation ofCdTe itself would be negligibly small.23 We note parenthetically that it might also provpossible to use AL E variant (ii). for example, with dimethycadmium and diethyl telluride as reactants, because thescompounds have been used to grow CdTe epilayers by metaorganic CVD.24 CdTe was also chosen for investigation because of ts potential importance in the areas of optoelectronics, integrated optics, and solar energy conversion. It is already used as a substrate material for Cd) xHgx Te solisolutions and as a constituent of CdTe-HgTe superlatticeswhich make excellent infrared detectors for the 8-14 pmatmospheric window. 25.26

It should be noted in passing that CdTe single-crystasubstrates of high structural perfection are difficult to obtain. Other substrates, such as GaAs (100), lattice-matching InSb (100),27.28 and (111),27 sapphire (0001),29 and S(111),30 all of which have been employed in the MBgrowth of CdTe, offer viable alternatives to CdTe substratefor growth of CdTe epilayers.

ALE experiments with nonpolar (110) CdTe substrates,20 polar (111) A and (111) B substrates,1,21.22 as weUas GaAs (100)31 have been carried out. In order to determine an appropriate growth temperature. an amorphous Teoverlayer about 1 nm thick was deposited onto the substratat room temperature. 20 The substrate was then graduallyheated to 800 K while observing the TeM 4.5 N 4,5 N 4,5 Augesignal at 485 eV. The intensity of this signal decreased withincreasing temperature up to 540 K (see Fig. 6) , abovewhich it leveled off. This change was due to the desorption oloosely bound Te atoms in the deposit, and the constan t signal in the range 550--750 K corresponded to the stable surface. Similar results were obtained for Cd. These and othesimilar experiments indicated that at 540 K thick amorphous deposits reevaporated at the rate of roughly 300 nmls



7/18

++i7I 55 + + +

+:.::4:LLJ>LLJ0::

50+ + + ++ + - r - - - + " ' - - ~ - - , o r l + + +

400 500 600 700SUBSTRATE TEMPERATURE (K)FIG. 6. Peak-to-pea.ic. height ofthe48S-eV Auger signal ofTe for a Te deposit on a CdTe (110) substrate as a function of substrate temperature.Loosely bound Te atoms desorb at high temperature. Auctuations in theAES peak intensities are mainly due to quick recording of the signals atshort intervals while heating the sample.

forTe and 27 pm/s for Cd, confirming that ALE variant (i)was indeed operative.

With the substrate at 540 K, Cd and 'Ie were alternatelydeposited as molecular beam pulses ofed and Te 2 The deposition rate was varied from 0.01 to 0.08 nm/s, as monitored by a quartz crystal oscil.lator. In order to obtain arough estimate of actual growth per cycle, films were deposited onto small pieces of molybdenum at 540 K. The Augersignals from Cd, 'Ie, and Mo were recorded as a function ofthe number of ALE cycl.es. After 10 cydes it was found thatthe MoMNN Auger line could no longer be detected. Fromthis it was deduced tha t a film of thickness between 2 and 2.5nm must have grown, in good agreement with the expectedgrowth per cycle of 0.229 nm, the (110) interplanar spacing.

LEED patterns of the (110) face of the grown filmsshowed a simple (1 Xl) surface unit cell. (This does notpreclude reconstruction; see Ref. 32). From LEED patternsand AES and XPS spectra clean and stoichiometric singlecrystal films could be obtained for T T of 480 K and above.At lower temperatures, e.g., 390 K, diffuse LEED spots suggested poor crystal structure and, usually, deviations fromstoichiometry. This result was expected because the vaporpressures of elemental Cd and Te are about 10-4 and 10-8Pa, respectively, so that some trapping of excess Te may haveoccurred.

It is thought that the ideal ALE model cannot fully account for LEED observations which indicate that a growingfilm may possess better crystal structure than the CdTe substrate.6,21 It was partly as a consequence of these findingsthat an intermediate surface adsorbed layer was suggested tobe present during deposition. Based on experimental. studiesof isothennal reevaporation rates of elemental Cd and Tedeposits, it was estimated that reevaporation rates ofC

depositionflux mree'Jtlporntion ~ t ; 6 - few monolo)'ersI of the depositf o ~ t ~ u r f o c e : next nearestmlgrntion . , : to the substrnfl!chemisorption I"/ urfacereevaporntion - t "" - -!irst monolayerslow surface --- ---- of the dePOSit. t - - - . / _ L ~ _ nearest to themlgrn Ion . 0T77?r/J./ r;npT7 substratecrystallization I / 'substrate' I /1, ' I

,/ / i, /

FIG. 7. Schematic iUustration of the key conceptual steps associated withgrowth of CdTe overlayers.

just above the substrate would allow ready migration andmixing of atoms and molecules. The activation energies forreevaporation of this near-interface region were measured tobe 1.5 eV forTe on CdTe (111)AandO.5 eV forCdonCd'Ie(I l l )B. Because the activation energies for surface diffusionare only about one-third of the bond energies,33 i.e., 0.5 and0.17 eV for Te and Cd, respectively, migration of the adsorbed atoms and molecules should, indeed, take place effectively in the near-interface region.

In order to study the interrelation of timing and dose ofthe pulses, growth experiments on CdTe were perfonned byvarying the beam fluxes, lengths of duration of the puisesand time windows while maintaining Tg r = 553 K constant. 7 After each growth. the surface morphology was studied by a scanning electron microscope. The smoothest surface was obtained when the growth parameters were"optimized.." As an example of this optimization at a relative!y low net growth rate the following set of parametersmay be chosen:

(a) growth rate of 0.12 nm/s of the Cd and Te deposits;(b ) length of duration of pulses from 2.5 to 2.8 s;(c) delay times (windows) 0.5 s for the Te2 pulse incident on CdTe (111)A and 1.0 s for the Cd pulseincident on ( 111) B.In an experiment, these parameters yielded growth of a

monomolecular layer of 0.37 nm per cycle in 7 s. corresponding to a net deposition rate of 190-200 nmlh.

Angle-resolved UPS. which is very sensitive to surfacecrystal structure and contamination, gave further infonnation.20 The emission peaks were sharper and of higher intensity for ALE-grown films than for the original substrate,suggesting improved quality and, probably, a smoother surface. In addition, the films grown under optima! conditionsshowed features that are not observed with bulk CdTe( 110)unless it is cleaved in vacuum. These features of the electronic band structure were interpreted as due to the presenceof occupied surface states caused by surface reconstruction.32 They could not be seen in films grown at temperaturesgiving diffuse l ~ E E D spots, nor for those exposed to air. sothat they were definitely associated with a rather perfect surface.CdTe overlayers on CdTe (111) B are often better incrystal structure than those on CdTe (11 I )A, or on CdTe(110), as deduced from LEED. I t is also noted that thesharpness of LEED patterns improves as films grew thicker,



8/18

a)

b)

FIG. 8. LEED patterns (a) for a CdTe ( 111 ) B substrate (cleaned by Ar+ion sputtering then annealing at 470 K) and (b) for a 37-nm CdTe (1 11)Afilm grown on this substrate by ALE at 540 K.

0) ALE CdTe (111)

b) MBE CdTe (111)

19 1 ~ 1 ~ 1.WPHOTON ENERGY leV)

FIG. 9. Photoluminescence spectra taken at 2 K from (a) an ALE-grownCdTe/CdTe (I l l) sample and (b) an MBE-grown CdTe/CdTe ( i l l )sample. These films were prepared under closely similar conditions in aCdHgTe MBE reactor at Tampere UniveTSity of Technology.

R71 J. Appl. Phys., Vol. 60, NO.3, 1 August 1986

at least up to 50 nm. An example ofthis behavior is given iFig. 8 for growth on ( 111 )B. However, LEED only samplelocal order and, therefore, small coherent areas 20-30 nmacross may produce reasonably good LEBD patternsLEEO is also rather insensitive to non periodic surface defects and dislocations. Unfortunately, high-energy electrodiffraction measurements (RHEED), capable of probinlarge areas, have not yet been reported in the literature foALE-grown films.

Low-temperature photoluminescence (PL) from a fewALE-grown CdTe (111)B overlayers has been measured.Figure 9 shows a 2 K PL spectrum from a l-,um CdTe filmgrown at Tgr = 550 K in a CdHgT e MBB reactor (a t Tampere University of Technology, the reactor is not describehere). There are two narrow lines just above 1.59 eV. whicdominate this spectrum. They are most likely associatewith radiative recombination of excitons bound to shalloacceptors (p type). 34 Their strength and sharpness providevidence that the film is of high electrical quality. We a re noable to make a definitive assignment to the broad featurranging from 1.54 to L56 eV, nor to the weak doublet centered at 1.57 eV. However, a possibility arises that the broafeature is due to band-edge emission.34 For comparison, PL spectrum from MBE-grown CdTe prepared under closly similar conditions is also shown in Fig. 9. The two spectrare very similar. Perhaps the only difference observed is thintensity of the broad feature which is about three timegreater for the ALE-grown film.

In the recent literature much attention has been paid tgrowth of CdTe on foreign substrates. The major limitationof employing CdTe crystals are due to the high costs anunsatisfactory grain-boundary st ructures. GaAs is an attractive alternative because large-area wafers of high structurperfection [dislocation densities ::::: (1-5) X 104 cm- 2 ] arreadily available at a modest cost. Despite the fal;t that CdThas a large lattice mismatch of 14.6% with GaAs it growperfectly epitaxially in the (100) (Refs. 35--38) and (111(Refs. 37, 39-41) directions on GaAs (100). The growtmechanism of CdTe on this substrate has been studied idetail by Otsuka eta/. 37 and Datta etal.42 When CdTe (100is on GaAs (100), the (7X7) CdTe superlattice cell is amost exactly commensurate with the (8 X 8) GaAs cell witthe two crystals being separated by a thin interfacial oxidlayer. If no such oxide layer exists the growth is likely toccur in the ( III ) direction. 37 Massive dislocations of complex line nesting struc ture with line densities of lOll cm-are formed in the CdTe/GaAs interface,36 due to the latticmismatch. However, the dislocation line density reducegradually as the film grows thicker. In the near-surface region ( ::::: 100 nm) of the overlayer of thickness of a few mcrometers the dislocation line density is of the same order omagnitude as in GaAs wafers.

We have grown CdTe overlayers3J of hicknesses fromto 4,um on n-type GaAs (100) substrates [orientedtowards (110)]. Since the growth time required for thesfilms would have been discouragingly long by ALE alonelayer of 1-3 J.Lm was first deposited by MBE at the rate o0.5-1 ,um/h, after which the growth was cont inued by ALat a net rate of 0.2 ,umlh. Figure 10 compares LEBD pa



9/18

a)

b)

FIG, lO, LEED patterns (a) for a GaAs (100) substrate after heating at850 K (no Ar+ion bombardment) an d (b) for a 3,3-Itm CdTe (100) filmgrown on this substrate at T., = 570 K. Primary beam energy = 45 eV.

terns for a GaAs (100) substrate and a 3.3-JLm CdTe overlayer. In this example growth has occurred in the (100) direction, as indicated by LEED and by a x-ray diffractioncurve shown in Fig. 11. Furthermore, a scanning electronmicrograph (SEM) of the film of Fig. 10 (see Fig. 11) showsthat the surface is smooth under medium magnification andcomparable to that obtained usually by MBE. A mUltipointAES analysis indicates that the film is uniform and stoichiometric all over the film area (o f about 0.5 em;1), being reproducible to within a few parts per thousand.2. Cadmium manganese telluride, Cd1-xMnxTe

This solid solution is a diluted magnetic semiconductor,also known as a semimagnetic semiconductor, whose subIattice is partly made up of the substitutional magnetic ions. Itsternary nature permits one to vary the energy gap, the effective mass, the lattice constant, and other physical parameters by alloying Mn.43-4 S I t has interesting magnetic, optical, and magneto-optical properties,46 for example, a giantZeeman splitting47 with exotic consequences.48

Manganese is an incongruently evaporating elementhaving a low vapor pressure of 1.8x 10 - 15 Pa, compared to1.3 Pa for Cd and 8.3 X 10 - 2 Pa for Te, at Tgr = 540 K. Onemight therefore expect it to be impossible to make use ofALE variant (i). This problem has been circumvented by

515 150Ekin (eV)

0)

li '!:::z:;:,tl )a::>-f-ViZf-Z

a::x22 28 34 40

R72 J. Appl. Phys Vol. 60, No.3, 1 August 1986

CdTe( 400)

52

bl

&49 -

FIG. II . (a ) Scanning electron micrograph.(b) AES spectrum. and (c) x-ray diffraction fora 3.3-/im CdTe (100) overlayer on GaAs (100)of Fig. 10.



10/18

I

~ - - e . t o - - ' - ~ - ( j o a l ~ D I N G ENERGy (eV)

(Q) FIG. 12. X-ray photoelectron spectra (a) for Cdo 7 Milo 3Te (II) overlayerand (b) fora CdTe (11overlayer. The inset shows precision measurements fthe Mn 2p core level emission from Cdo7MIlo.3 Te anfrom a metallic Mn overlayer grown on this compoun_ _.J - - Cd

1054KINETIC ENERGY (eV)

careful control of the amount of Mn during deposition. Twodifferent methods were used22 ;(a ) The pulsed beams of Cd and Mn were generatedsimultaneously by opening and closing the effusion cell shutters at the same time, with the relative Cd and Mn beamintensities set at a fixed value (e.g., 10: 1);

(b) The beam intensities of Cd and Mn were equalizedbut the duration ofthe Cd pulse was made much longer thanthat of the Mn pulse (e.g., by a factor of 10).It is then clear that the mole fraction x is determined by theintensity and the length of duration of the Mn pulse. Aninteresting feature in this stepwise growth is that when theMn pulse is adjusted to produce a coverage less than one fullA73 J. Appl. Phys Vol. 60. No.3. 1 August 1986

340

atomic layer, a fixed value of x < 1 is always obtained pcycle; excessively arriving Cd and Te2 or variations in thefluxes (within certain limits) do not alterx. On the contraMBE-type evaporation results in continuous layer growthbeing determined by the relative concentration of Mn, Cand in the vapor phase near the sample surface; avariation in the beam fluxes is reflected in x. We should nohowever, that this particular AL E variant cannot be usedgrow any larger film than that obtained by MBE because opening of the cones of molecular beams is limited by gemetric factors similarly to ALE and MBE.

Cd'_xMnx Te films with x ranging from 0 to 0.9 habeen successfully grown by AL E on CdTe ( 111 ) B substraC. H. L. Goodman and M. V. Pessa A

Downloaded 29 Nov 2011 to 210.212.129.125. Redistribution subect to AIP license or co ri ht see htt ://a .ai .or /about/ri hts and ermissions


11/18

at T r = 540 K.22 The pulse length was varied in differentexperiments from 10 to 20 s for Cd and Te2 and from 2 to 20 sfor Mn. The beam intensities were chosen to correspond tothe deposition rates of 0.05-0.10 nm/s for Cd and Te2 and0.01-0.04 nm/s for Mn. In all cases less than a monolayer ofMn was deposited in each cycle.

The Cd -x Mnx Te films were found to retain essentiallytheir zinc-blende structure up to at least x = 0.9, in sharpcontrast to bulklike crystals. We would like to note in thiscontext that even pure MnTe overlayers of a zinc-blendeform, when grown on appropriate substrates, are conceivable. Furdyna49 suggested that there exist wen latticematched substrates, such as Cdo.6sZIlo.35 Te andInSbo.77 PO23 on which MnTe of the zinc-blende structuremight be anchored.

Little work has been done so far on spatial distributionof magnetic ions of semi magnetic semiconductor films. Oneway of studying this question is to measure x-ray photoelectron spectra of the films. Figure 12 shows MgKa-XPS ofpure CdTe ( I l l ) and Cdo.7MIlo.3 Te(11!) grown on CdTe( 111 )B. I t can be seen that upon alloying Mn a large reduction in relative intensity of the Cd 3e3/2. 5/ 2 doublet occursand the Mn 2P1/2. 3/2 lines appear. High-precision measurements shown in the inset of Fig. 12, where XPS spectra of theCdo.7 MIlo.3 Te overlayer and a 2-nm metallic Mn overlayerdeposited onto the Cd{).7MIlo.3 Te are compared with eachother, provide closer insight into substitutional disorder ofthe sublattice. The chemical shift of2.6 eV of he leading Mn2PS/2 peak, the line broadening, the presence of a satellitestructure, and the absence of a metallic Mn signal in thecompound spectrum provide evidence that manganese is indeed a constituent of random ~ . 7 MIlo.3Te with no significant metallic clusters.

A LEED pattern for an ALE-grown Cdo.82 MIlo. 18 Te( I l l) epilayer of 30-nm thickness on CdTe (111) is shownin Fig. 13. The surface is well ordered, exhibiting sharp(1 Xl) LEED spots with low background intensity.Figure 14 shows a schematic real-space energy band diagram of an ALE-grown unintentionally doped CdTe-C ~ . 6 MIlo.4 Te (111) multiquantum-weU heterostructure.Electronic valence-band features of these two componentLayers were studied by measuring angle-resolved UV photoemission. Apart from differences in details of emission fromthe valence bands no measurable valence-band offset wasobserved on alloying Mn (for determining band off-sets byelectron spectroscopy, see e.g., Ref. 50). Thus the difference

FIG. 13. LEED pattern of a 30-nm Cdo.S2 MIlo.,sTe (111) film depositedonto a 37-nm (I l l) buffer layer on CdTe.R74 J. Appl. Phys., Vol. 60, No.3, 1 August 1986

Cd.6 Mn lo Te-1 __ - -

o , 2 3 4 5 6~ ~ T l i T .

LAYER THICKNESS (nm)

FIG. 14. Schematic real-space energy band diagram of a Cdo.6MIlo . Te-CdTe multiquantum-well heterostructure grown by ALE (upper panel).The valence-band maximum (VBM) remains unaffected on alloying Mn towithin an experimental accuracy of 50 meV investigated by angle-resolvedphotoemission; the difference in band gaps influences the conduction-bandminimum (CBM) alone. The lower panel shows a study of the interface atthe position indicated by the vertical line. This stud y was based on measurements of the peak-to-peak intensity ratios of Cd MNN, Mn LMM, and TeMNN Auger transitions; red = [CdI lT. and R Mn = [Mn/ IT. The solidcurves are calculated assuming that the interface is sharp and growth occursin a layer-by-Iayer fashion, one molecular layer pe r cycle.

in the band gap Eg at the interface affects only the conduction-band minimum, as illustrated in Fig. 14.The CdTe/Cdo.6MIlo.4 Te (111) interface was studiedin a nondestructive way by combining the ALE and AEStechniques. Very thin overllayers of CdTe were deposited

onto Cdo.6 MIlo.4 Te, and Mn LMM, Cd MNN, and Te MNNAES signals were measured at room as a function of overl.ayer thickness. The results are shown and. compared withtheoretical estimates for relative signal intensities in Fig. 14.The solid curves represent the calculated behavior of CdMNN and Mn LMM peak-to-peak intensities relative to TeMNN peak-to-peak intensity, R Cd = I Cd / I To andR Mn = I Mn /1 Te' respectively, on the assumption that thejunction is sharp. fn other words, no interdiffusion ofMn orother grading effects are assumed to occur in the interfaceregion. Furthermore, the well-known law of exponential decay of low-energy electron signals in solids was applied. Theinelastic mean free paths of Cd MNN and Mn LMM electrons were taken to be 1.4 and 1.9 nrn, respectively. Judgingfrom the measured and calculated behavior of the intensity

C. H. L. Goodman and M. V. Pesss R74Downloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions


12/18

ratios one may conclude that the interface is extremelysharp, possibly of an order of atomic diameter. For comparison, by the standard AES-ion milling technique51 the interface region of MBE-grown CdTe/Cd l _ x Mnx Te heterojunctions with x ranging from 0.20 to 0.53 has been found tobe as large as 10 nm. Such a graded junction is an artifactwhich arises from an intermixing and cratering of the junction region during ion bombardment.3. Gall/um arsenide, GaAsNishizawa et a/. 9 11 have described the growth of GaAsusing ALE variant (ii). Trimethyl gallium (TMG) andAsH3 were cycled alternately over a single-crystal substrate,using the reaction

Ga(CH3 )3 +AsH3--GaAs + 3CH4The apparatus used is shown in Fig. 15. Arsine was admittedat a pressure of 10- 4_10- 5 Pa for times of 2-200 s thenevacuated. A s tandard cycle of AsH3 20s, evacuated for 3s,TMG 4 s, evacuated for 3 s (20,3,4,3) was adopted (seeFig. 16). AsH3 pressure was usually 7x 10 - 3 Pa, TMG admittance was varied. I t was found possible to obtain ALEgrowth at temperatures between 720 and 970 K.The thickness of deposit grown per cycle d varied quitemarkedly with the amount ofTMG admitted to the apparatus in each cycle. From the initial br ief publicationII d saturated at about 0.55 nm for Tgr = 870 K, as shown in Fig. 17.This corresponds to about two layers of GaAs per cycle on( 100). As would he,expt.'(:ted for the ALE mode of growth, dremained c o n s t ~ n l m:e! large numbers of cycles, as shown inFig. 18 which cor.msponds to a TMG dosage of about 10- 2Pa 1/s, and ad value of0.4 nm. While the constancy appearsto confirm that ALE does occur, the variation of d with theamount of TMG admitted per cycle lUust presumably indicate that even an incomplete coverage can lead to growth inthe ALE mode, while, as already noted, sufficient TMG canallow up to two GaAs (100) layers to form per cycle eventhough this is not too readily explained on the simple theorypresented in Sec. n t In their later work Nishizawa et aUfound much smaller d values by lowering the growth tem-

Light 1\

= = : ; : = = t = = = l P y r o 2 / m e t ~ QuartzPlate ...........

I\ I\ I

A S H l ~ > < 3 = = l = = = " \ I_ ~ > < ) : * = = \ ubstrateTHG

-QMS

Pedestol

Pumping System

FIG. 15. Schematic drawing of equipment for atomic layer epitaxy ofGaAS."lI Lights I and 2 are for photoexcitation (see text ). The lamp is forheating the substrate (QMS stands for a quadrupole mass spectrometer).R75 J. Appl. Phys., Vol. 60, No.3, 1 August 1986

10 Admittance Mode (20': 3';4': 3")

o JMGo 40 80 120 160 200Time (sec J

FIG. 16. Gas pressure in growth chamber of Fig. 13 during standard ALcycles (20, 3,4,3 s).

peratures and increasing the TMG dosage. At Tg r = 770 Kd was found to saturate at a value very slightly below thO.283-nm characteristic of a true monolayer of GaAs o(100); i.e., the simple ALE growth model was wen obeyed

One problem with the reaction betweenTMG and AsHis that its rate falls off rapidly below 870 K apparently because of the energy required to dissociate arsine. Nshizawa

52showed that it was possible to use UV radiatiofrom a high-pressureHg lamp or an excimer laser to alleviatthis source of difficulty. Growth could be obtained down ttemperatures as low as 560 K. This appears to be a truenhancement of reaction by the radiation, and not an effecdue to heating-any temperature rise of he substrate causeby the radiation would be no more than about 1 K. It waalso found that below 870 K the growth thickness per cycdecreased quite sharply for constant TMG dosage. Even very low d values, however, a constant growth per cycle wmaintained.Some electrical results have been quoted by Nishizaw

et aU The films produced at low temperatures, - 620 Kwere p type, with hole concentrations of the order of 10cm- 3 This high defect concentration was attributed to thincorporation of carbon. Some improvement on this was otained by using triethyl gallium, (TEG) which stilJ gave ptype material but with carrier concentration of the order10 18 cm- 3 Most recently, Nishizawa52 reported on carriconcentrations below the 10 17 cm -3 level. No results hav

0.6

0.5ec:: 0.4.!!!u0.3a;C>.:: 0.2:kel!J 0.1 Tgr = 870 K

0 0 2 3TMG Admittance Quantity per Cycle (Torr Vs)

FIG. 17. Thickness per cycle ofGaAs as a function ofTMG admittance

C. H. L. Goodman and M. V. Pessa RDownloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions


13/18

'"'"'"-'"',! 1

0.4 nm/cycleTgr = 870K

1000 2000 3000 4000 5000Number of Gas Admittance Cycle

FIG. 18. Thickness ofGaAs grown at 870 K vs number of ALE cycles (20,1,4, Is).

been reported for material grown at higher temperatures, orfrom higher alkyls e.g. tri-isobutyl gallium, either (or both)of which might lead to still lower carbon incorporation.Usui and Sunakawa10 at Nippon Electric Corp., Japan,have also prepared GaAs by ALE. They employed a dualgrowth chamber reactor in hydride vapor-phase epitaxy.The principle of his growth system is illustrated in Fig. 19.The substrate was exposed alternately to GaCI and AS 4 byswitching its position between the two chambers (the AsH3vapor stream was interrupted during the substrate transfer).The period of the cycle was 60 s but considerably shortercycles could also be applied.

The growth thickness per cycle for several substrate surfaces was studied as a function ofHCl(Ga) flow rate at Tgrof 820 K. The d values thus obtained remained almost constant independent of the flow rate in the range from I. to 5cm3/min and closely corresponded to the thickness of amonolayer on each surface (see Fig. 20).

The films were monocrystalline, as revealed by reflection high -energy electron diffraction (RHEED) measurements, with completely mirrorlike surfaces (Fig. 21). Nooval defects typical ofMBE-grown GaAs were present. Furthermore, it is believed that ALE growth reduces the number of EL2 deep traps, but this has not yet been demonstrated.

Of particula r importance was the observation10 that selective growth at windows of Si02 masks on GaAs did no tshow any significant enhancement in d. This result impliesthat ALE might be suitable for selective epitaxy.I t is also interesting to note that preparations for growthof III-V mixed crystals by ALE are under way at NEe. Preliminary results on InGaP have already been obtained. Theyare found to be very encouraging. 10

ALE growth of GaAs and AlAs from metalorganic(TMG and TMA, respectively) and hydride sources has

H2, AsH3 J AS4 IU=== :JH2 Hcn Go GaCl -" r...------ I- \. -T t:.r------FIG. 19. Schematic diagram of the dual growth chamber reactor used forALE growth of GaAs. 0 The substrate is alternately exposed to the tworeactants (figure courtesy of A. Usui).R76 J. Appl. Phys., Vol. 50, No.3, 1 August 1986

0.5GaAs

E 0.4 (211) ." (111)(511)-g. 03 (100) (100)... (111)A 0'".J;; 0.2 (110)"il!J 0.1

00.1 10

HCIIGa) fcc/min]FIG. 20. Growth rate per cycle for GaAs overlayers as a function of flowrate of He!. Tg, = 820 K (figure courtesy of A. Usui).

a)

SURFACE PHOTOMfCROORAPHb)

FIG. 21. (a) Photograph ofa GaAs (100) film grown by ALE. 0 (b) Scanning electron micrograph of he film exhibiting no oval defects (figure courtesy of A. Usui).

c. H. L. Goodman and M. V. Pessa R76Downloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions


14/18

Fixed ar tRecess for substrateRotating part

Quartz tube

ExhaustFIG. 22. Schematic diagr am of the growth chamber and susceptor of ALE(Ref. 53) for preparing GaAs and AlAs. The susceptor consists of a fixedpart, a rotating part, and a recess (in the rotating part) which holds thesubstrate. The H2 flow in the middle tube helps prevent mixing of the reactants.

been reported by Bedair et al.53 Their growth chamber (seeFig. 22) has a simple, ingeniously designed structure whichmay permit one to grow several samples at the same time in acyclic fashion. The susceptor incorporates a shuttering action which allows successive exposure to streams of gasesfrom two sources, viz., from AsH3 + Hz and TMG + Hz forgrowth of GaAs, or AsH3+ H2 and TMA + Hz for growthof AlAs. A large flow of Hz in the middle tube is designed toprevent mixing of the reaction gases from the adjacent tubes.When the substrate is rotated away from the growth (window) position by a rotating part, most of the boundary layerwill be sheared off by a fixed plate placed above the rotat ingpart facilitating an almost immediate termination of exposure of the substrate to the input flux. In the experiments,one complete cycle was performed in about 10 s; Tgr forgrowth of GaAs and AlAs was in the range from 830 to 870K.

The GaAs and AlAs overlayers prepared. in this way onGaAs (100) substrates were single crystals, as deduced fromtransmission electron micrographs. Photoluminescencemeasured at 77 K of 100 cycles of ALE-grown GaAs sandwiched between layers of MOCVD-grown GaASo.97 PO.03showed. a full width at half-maximum of 11 meV, indicatingthat the sample was of a good electrical quality.8. Poiycrystailine and amorphous deposits

Most of the published papers on ALE are concernedwith polycrystaUine and amorphous deposits. These arisemainly from the use of amorphous substrates. In part, poorcrystal structure may be due to carrying out deposition attemperatures that are only a small fraction of the absolutemelting point of the material concerned, so that surface mobility of adsorbed species would be small or negligible. It ispossible that many materials in this section could be grownR77 J. Appl. Phys., Vol. 50, No.3, 1 August 1986

as epitaxial layers if single-crystal substrates and highegrowth temperatures were used.

The various materials grown as polycrystalline or amorphous films using ALE cycling will be listed in tum.1. Zinc sulfide

This is the most widely studied material for ac EL thinfilm displays. The growth and properties of thin films ofZnand ZnS:Mn fabricated by ALE variant (ii) have been thsubject ofmany investigations (see, for example, Refs. 2, 5463. Either zinc chloride and hydrogen sulfide via the sequential exchange reaction:ZnC1 2 + H 2S--+ZnS + 2HCI(600 < Tgr < 800 K), (1

or anhydrous zinc acetate and hydrogen sulfide54 via threactionZn(CH3COOh + H 2S

_ZnS + 2CH3COOH(520 < Tg r < 620 K) (2can be used. The occurrence of reaction ( 1 ) was demonstraed by straightforward in situ AES measurements4 (see Fi23). The spectra shown were taken for a glass substrawhich was exposed alternately to pulses of ZnClz and Hunder UHV conditions.The reactions were found to leave ntraces of Cl or othe r impurities within the detection limitAES. By contrast, ZnS grown in a gas-flow reactor of Fig.from the same reactants exhibits traces of Cl and C.

Tammenmaa et a/.54 used reaction (2 ) to grow ZnS oglass. The maximum growth rate achieved was 0.26 nm pcycle, which corresponds to 6/7 of the lattice spacing0.312 nm in the preferential (111) growth direction. Theproposed that Zn(CH3COO)z forms multinuclear compounds in the vapor phase. Chemisorption of these specion the substrate surface leads to high coverages and, accoringly, to the relatively high growth rate observed.

W"CJ---"CJ

@ ~ KINETIC ENERGY lev

FIG. 23. Auger electron spectra showing an occurrence of stepwise growofa ZnS film on a glass substrate via surface exchange reactions. The spectwere taken at two successive phases of a deposition cycle. The wide-ranscan ( 100-1100 e V) in the middle was recorded a fter exposing the substra10 a ZnCl2 pulse; Cl is chemisorbed on the surface as indicated by the apearance of the Cl AES peak. The uppermost spectrum (100-200 eV) othe righ t-hand side shows the absenceofCI after a pulse ofH2S which completes the cycle, yielding a growth of a ZnS molecular layer. Oxygen aboron signals originate from the glass substrate.

C. H. L. Goodman and M. V. Pessa ADownloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions


15/18

Tammeomaa et al. 54 have also carried out doping experiments in connection with growth of ZnS via reaction(2), using volatile complexes of Eu, Th, and Tm. Photoluminescence measurements at room temperature showed that1b+ 3 ions gave a clear PL spectrum with green emissionpeaking at a wavelength of 547 nm. Intensity was found toincrease with Th concentration up to 1.5 at. %. In comparison with the yellowish-orange emission (peaking at A;:::; 580nm) from Mn+2, the most common activator in ZnS emis-sion from Th +3 was very small. 'The microstructure of ALE samples from Lohja, observed by the TE M technique exhibited no fine-grained region at any thickness.64 Fig. 24 shows a typical cross sectionof ZnS layers deposited onto amorphous A120 3 Very pronounced columnar grain growth occurred with many grainsextending from the bottom to the top of the layer. Selectedarea x-ray diffraction curves56 S8 showed that the films had acubic or hexagonal polytype structure for growth in the temperature range from 300 to 570 K. The specific orientationswere (111) for the cubic structure and (00.2) for the hexagonal structure which consisted predominantly of a 2H phase.The hexagonality increased remarkably in the range from630 to 770 K. The degree of preferred orientation of microcrystallites at Tgr = 770 K was so high that one-half of the(00.2) poles were aligned within 7 of the substrate normal.

Th e substrate material was also found to influence thetexture. The highest degree of orientation was obtained withfilms deposited onto a amorphous Ta20 s (o r A1 20 3 ) bufferlayer of hickness of about 50 om grown on a glass substrate.Insertion of a microcrystalline Sn02 layer between the glassand the Ta20s (o r AI20 3 ) caused some reduction in the degree of preferred orientation.

The crystaHite size varied as a function offi lm thickness.Typically, for films thicker than 200 nm prepared atTgr = 770 K, the length of coherently diffracting domains inthe direction normal to the surface ranged from 40 to 120 nmwith a size distribution maximum at 100 om. It isnoteworthy that the mean grain diameter parallel to the filmplane seen by TEM (Fig. 24) was also about 100 nm. Thisagreement between x-ray diffraction and TEM suggests thatthe columnar grains are made up of single crystals mainly ofa hexagonal structure with preferential (00.2) orientationobtained at high Tgr .

ZnS films grown from zinc acetate at Tgr = 5@-650 K

O.SlIm

FIG. 24. Transmission electron micrographs of ZnS:Mn films grown onamorphous A120 , by ALE at Lohja Corp. (a ) Corning 7059 glass substrate'(b) soda lime glass substrate. 'R78 J. Appl. Phys., Vol. 60, No.3, 1 August 1986

FIG. 25: Half-page matrix display module for personal computers. Thehght-emlttmg layer of ZnS:Mn is sandwiched between two dielectric layersof mixed AlzO,-Ti02 grown by ALE. The transparent front electrode isITO made by sputtering. The back electrode can be either ITO or alumi? u ~ . The structure is protected by a passivation layer of AI 20, ; black printmg IS appbed on top of the passivation layer to increase contrast.

exhibited the average crystallite sizes of 40-80 nm, depending on the film thickness. 63 E!ectrorefl.ectance studies65 indicated that two crystal phases, cubic and hexagonal, coexistin these films with the cubic structure predominating at thelower end of the Tgr range.

I t is thus concluded that crystal structure of ZnS filmsdepends both on growth temperature and on source materials. The substrate has also an influence on the specific orientation of crystallites.

The dislocation densi ty in ZnS films grown at 770 K wasfounds8 to be about 1010 cm - 2 . This is an order of magnitudelower than the dislocation density in ZnS grown by electronbeam evaporation.6O

The ZnS:Mn films used by Lohja in their ac thin-filmEL displays (see Fig. 25) exhibit a maximum luminance of3000 cd./m2 at 1. kHz. An EL external efficiency of about 2ImlW obtained at 1 kH z has been observed.66 At very highfrequencies the external efficiency may rise up to 8 Im/Wwhich is probably the highest efficiency attained for activelayers of ZnS:Mn grown by any technique. These displaysalso require the introduc tion of dielectric layers, e.g., A1 20 3,again grown by ALE. Uniform pinhole-free deposits of ZnSand dielectric layer materials can be obtained over very I.argeareas-250X 150 mm 2 glass substrates are routinely used atLohja.

2. Zinc tellurideZnTe films grown from elemental Zn and Te exhibit

smooth and featureless surfaces. X-ray diffraction measurements revealed that the films tended to crystallize preferentially in the (111) direction. S An interesting observation isthat in a comparison of x-ray diffraction curves of the filmsgrown in UHV by MB E and AL E as a function of Tgr' theALE-grown films show a weaker temperature dependenceof the texture than those grown by MBE. That reasonablyclose control of stoichiometry was maintained during ALEgrowth could be deduced from XPS and AES spectra.4



16/18

3. OxidesZinc acetate and water vapor have been used as reac

tants for growing ZnO.54 .59The following reaction appearedto hold:Zn(CH3COOh + lizO

..........Zn O + CH3COOH (560 < Tg r < 630 K). (3 )The growth rate was typically 60 nm/h, and thus one-fifth ofthe growth rate observed in most cases for ZnS. X-ray diffraction revealed that the ZnO films were of rather poorcrystallinity. Presumably this was because ZnO is a morerefractory material than ZnS, so that the surface mobility ofatoms at 600 K would be rather small (see comments atbeginning of Sec. IV).

Growth ofTazOs by ALE variant (ii) has been realizedvia the reaction

2TaCls + 5 H z O " " " " " T ~ 0 5 + lOHC} (Tgr z770 K) . (4)That this exchange reaction in fact occurs has been demonstrated by in situ AES measurements4 of a substrate whichwas exposed alternately to vapor pulses of TaCls and H20 .Cl was seen in AES after each pulse ofTaCls but was alwaysremoved by a subsequent pulse of H20. T ~ 0 5 layers areamorphous (again Ta20 S is a refractory material with a veryhigh melting point). It should be noted that the extremelylow vapor pressure ofTa at 770 K would prevent ALE variant (i) being employed.

Amorphous layers ofA1 20 3 are grown from anhydrousAlC13 and water vapor at Tgr :::::: 720 K. They are pinhole-freeand can therefore be used as an ion barrier and dielectric andpassivation layers in EL devices. Although A120 3has a reasonable dielectric strength by itself, essential improvementhas been achieved using a mixture of A120 3 and TiOz, prepared by ALE. SnOz films are grown from SnC14 and watervapor. They are noted as exhibiting some weak x-ray diffraction peaks, and so are not completely amorphous.

Work is also being done at Lohja67 on indium tin oxide(ITO) used as a transparentconductor in displays. Resistivityof 1.6X 10 - 4 n cm, with transmission of more than 80%has been reproducibly demonstrated over a large substratearea.

V. THE EXTENSiON Of ALE TO OTHER MATERIALSIn principle it should be possible to extend ALE to a

wide range of materials, even those in which one of the constituents is in volatile, since there appears to be a general family of reactions that might be used for achieving AL E variant(ii). These can be summarized:Metal halide + H 20 or oxygen .........Metal oxide;Metal halide + H2S or sulfur vapor.........Metal su:lfide;Metal halide + H 2Se or selenium vapor .........Metal selenide.

One could well expect that closely related compounds tothose discussed in the preceding section (ZnC12, AICl3,SnC14 , and TaCls), should behave in very much the sameway, for example:CdClz, HgCh, GaCl3, Inel3, SiC14, GeCl4,ZrC14, NbCls, and also other transition metal halides, e.g.,Feel3 In addition, the success of Nishizawa and Bedairet a/.in using TMG,9.II,S3 TEG,9.11 and TM A (Ref. 53) in theR79 J. Appl. Phys., Vol. 50, No.3. 1 August 1986

preparation ofGaAs and AlAs (Ref. 53) suggests that metalorganies generally might be utilized as metal sources. Somcaution may be needed here, however, no t so much on account of the toxicity and pyrophoric nature of metalorganies, but with regard to their thermal stability. That must bsufficient to withstand an encounter with the ho t substrawithout decomposition in order that desorption of physsorbed material would occur.

It is also important to consider the second reaction component. On general thermodynamic grounds oxygen or sufur will not react with a chemisorbed halide layer at so Iowtemperature as water vapor or "2S, since such reactiowould require greater free energy for the halogen to be giveoff as an elemental vapor than as a hydride. Since there is major drive with electronic materials towards the use of thlowest possible temperature for material preparation, thcould favor the use of the hydrides rather than the pure elements. A further important point is that H 20 , HzS, etcgreatly resemble H3N, HJP, H 3As, and, as Nishizawa's worwith GaAs has shown, it should be possible for such molecules to react with chemisorbed metal compounds to faciltate ALE growth of nitrides, phosphides, arsenides, etc. Thfirst successful growth of phosphides (InGaP) has alreadbeen carried out. 10

It should also be possible to prepare more complex materials than the relatively simple compounds discussed sfar. Perrites would serve as an example to illustrate thpoint. Nickel zinc ferrite could in principle be prepared fromNiCI2 , ZnC12, and FeCl3 or manganese ferrite from MnC1and FeClz. The magnetic properties of ultrathin films osuch materials remain essentially unknown and would warant attention. I t could, for example, be interesting to prepare multilayer structures of ferrites combined with nonmagnetic analogues in order to obtain direct access to effecof strain and "two-dimensionality" on magnetic propertieFor development of magnetic properties, layers of good crystal perfection would be required. In order to obtain themALE would need to be carried out on a substrate held at higtemperatures, say, 0.5 of the melting point. In the case omost ferrites this might be 1000 K or more. No experimenon the ALE growth ofoxides at such high temperatures hayet been attempted.

Finally the ALE approach has so far been restricted tcompounds. It is conceivable that elemental materials, e.gsilicon, could be grown via chemisorption of a monolayer osay, a chlorosiJane from a stream of argon, followed breduction of the chlorosilane in a pulse of hydrogen.VI. CONCLUSiONS

Crystal growth in the AL E mode, whether by physicor chemical vapor deposition, is necessarily slow becauindividual layers of atoms are deposited with a time of thorder of one second being required for each layer. Althougit seems unlikely that such times could be shortened vesignificantly there are advantages, e.g., the feasibility of dpositing several large-area films simultaneously, whicmake long growth times acceptable.

Where ALE might offer particularly striking advatages is in the growth of ultrathin epdayers under preci

c. H. L Goodman and M. V. Pessa RDownloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions


17/18

and, almost, automatic control. The technique gives a"growth per cycle" rather than a growth rate, and does so atrelatively low temperatures. It therefore should enable heterojunctions with abrupt interfaces, quantum well structures and superiattices to be grown without the need for thecomplex and costly control systems. Additionally, ALE mayprovide a vehicle for investigating fundamental aspects ofcompound semiconductor growth. However, i t still remainsto be demonstrated that the perfection of structure and control of mpurities and stoichiometry required of materials forelectronic devices can in fact be achieved.With regard to polycrystalline and amorphous materials, the ac thin-fi]m EL displays made by the Lohja grouphave demonstrated unequivocally that ALE can be used togrow highly uniform deposits of good structural integrityboth mechanicalJy and dielectrically speaking. These qualities are of importance in many other fields, too. Applicationto ternary or higher compounds could also prove feasible,and low-dimensional magnetic structures might eventuallybe fabricated by ALE.ACKNOWLEDGMENTS

The authors wish to thank Dr. A. Usui (NEC) and Dr.M. Tammenmaa (Helsinki University of Technology) forproviding us with their results prior to publication. Thiswork was supported in part by The Academy of Finland andThe Center of Technological Development (Finland).

IT. Suntola and J. Antson, Finnish Patent No. 52359 (1974) and T. Suntola and J. Antson. US Patent No. 4058430 (1977); T. Suntola, A. Pakkala, and S. Lindfors, US Patent No. 4389973 (1983).

'T . Suntola, J. Antson, A. Pakkala, and S. Lindfors, Soc. Information Display (SID'80) Dig., 109 (1980).3M. Pessa, in Optoelectronic Materials and Devices, edited by M. A. Herman (PWN-Polish Science Publishing, Warsaw, 1983), p. 217 and inExtended Abstracts of the First Symposium on Atomic Layer Epitaxy,Espoo, Finland, Dec. 13-14, 1984 (unpublished).

4M. Pessa, R. Makela, and T. Suntola, Appl. Phys. Lett. 38,131 (1981).5M. Ahonen, M. Pessa, and T. Sunto1a, Thin Solid Films 65,301 (1980).6M. Pessa, O. Jyl ha, and M. A. Herman, J. Cryst. Growth 67, 255 (1984).7M. A. Herman, O. Jylhil, and M. Pessa, to be published in Crystal Research and Technology ( 1986); for applicationof he generalized model ofALE variant (i ) to growth of CdTe, see M. A. Herman, M. Vulli, and M.Pessa, J. Cryst. Growth 73, 403 (1985).

"T. Pakkanen, M. Lindblad, and V. Nevalainen, in Extended Abstracts ofThe First Symposium on Atomic Layer Epitaxy, Espoo, Finland, Dec.13-14,1984 (unpublished).

9J. 1. Nishizawa, H. Abe, and T. Kurabayashi, J. Electrochem. Soc. 132,1197 (1985).lOA. Usui and H. Sunakawa, The 12th International Symposium on Gallium Arsenide and Related Compound, Karuizawa, Japan, Sept. 23-27,1985 (unpublished); A. Usui (private communication).

IIJ. I. Nishizawa, International Conference on Soli d State DevicesandMate-rials, Kobe, Extended Abstracts (Japan Society of Applied Physics, Tokyo, 1984), p. l.

12K. A. Jackson, in Progress in Solid State Chemistry, edited by H. Reiss(Pergamon, Oxford, 1967), Vol. 4, Chap. 2, pp. 53-SO.

13p. Hartman, in Crystal Growth: An Introduction, edited by P. Hartman(North-Holland/American Elsevier, AmsterdamlNew York, 1973),Chap. 14, pp. 366--402.

141. Daweritz, Surf. Sci. 160, 171 (1985).151. B. Theeten, 1. Hollan, and R. Cadoret, in Crystal Growth and Materi-

als, edited by E. Kaldis and H. Scheel (North-Holland, Amsterdam,1977),Chap. 16,pp. 196-235.

R80 J. Appl. Phys., Vol. 50, No.3, 1 August 1986

16D. W. Shaw, Crystal Growth, Theory and Techniques, edited by C. H. L.Goodman (Plenum, New York, 1974), Vol. I. Chap. I, pp. 1-48.

"C . Engler and W. Lorenz, Surf. Sci. 104, 549 (1981).18J. Korec and M. Heyen, J. Cryst. Growth 60, 297 (1982).I"'N. Lorenz and e. Engler, Surf. Sci. 114,607 ( 1982).20M. Pessa, P. Huttunen, and M. A. Herman, J. Appl. Phys. 54, 6047

(1983).21M. Pessa, O. Jylhii., P. Huttunen, and M. A. Herman, J. Vac. Sci. Technol.I A 2, 418 (1984).22M. A. Herman, O. Jylhii., and M. Pessa, J. Cryst. Growth 66, 480 (1984);

M. Pessa and O. Jylhii., Appl. Phys. Lett. 45, 646 (1984).23D. L. Smith and V. Y. Pickhardt, J. Appl. Phys. 46, 2367 ( 1975).241. Bhat, L.M.G. Sundaram, J. M. Borrego, and S. K. Ghandhi, in Ab

stracts of 1984 Fall Meeting of Materials Research Society, Boston, Mass.Nov. 26-30, 1984 (unpublished).

25M. A. Herman and M. Pessa, J. Appl. Phy s. 57, 2671 (1985) and references therein.

Z6J. P. Faurie, A. Million, andJ. Piquet, Appl. Phys. Lett. 41,713 (1982);J.P. Faurier, in Extended Abstracts of The 1984 US Workshop on ThePhysics and Chemistry of Mercury Cadmium Telluride, San Diego, Ca,May 15-17,1984 (unpublished).

z7R. F. C. Farrow, G. R. Jones, G. M. Williams, and I. M. Yong, App!.Phys. Lett. 39,954 (1981).

28R. F. C. Farrow, A. J. Noreika, F. A. Shirland, W. J. Takei, S. Wood, J.Greggi, and M. H. Francombe, J. Vac. Sci. Technol. A 2,527 (1984).'''T. H. Myers, Y. Lo, R. N. Bicknell, and J. F. Schetzina, Appl. Phys. Lett.42,247 (1983).

3y. Lo, R. N. Bicknell, T. H. Myers, and J. F. Schetzina, J. AppJ. Phys. 54,4238 (1983).31M. Pessa and O. Jylha (unpublished work).He. B. Duke, A. Paton, and W. K. Ford, Phys. Rev. B 24, 3310 (198 I ); J.

Vac. Sci. Technol. 20, 778 (1982).33J. A. Venables, Vacuum 33, 701 (1983) ; J. A. Venables, G. D. T. Spiller,

and M. Hanbucken, Rep. Prog. Phys. 47, 399 (1984).340fhe PL spectra were measured and partly interpreted by B. Monemar at

The University of Linkoping, Sweden.3'K. Nishitani. R. Okhata, and T. Murotani, J. Electron. Mater. 12, 619

(1983 ).30R. N. Bicknell, R. W. Yanka, N. C. Giles, J. F. Schetzina, T. J. Magee, C.

Leung, and K. Kawayoshi, Appl. Phys. Le tt. 44, 313 (1984).37N. Otsuka, 1. A. Kolodziejski, R. L. Gunshor, S. Datta, R. N. Bicknell,

and J. F. Schetzina, Appl. Phys. Lett. 46, 860 (1985).38J. J. Dubowski, D. F. Williams, P. B. Sewell, and P. Norman, App!. Phys.

Lett. 46, 1081 (1985).39J. T. Cheung, M. Khoshenevisan, and T. J. Magee, Appl. Phys. Lett. 43,462 (1983).'"'P. P. Chow, D. K. Greenlaw, and D. Johnson, J. Vac. Sci. Technol. A 1,562 (1983).

4IH. A. Mar, K. T. Chee, and N. Salansky, Appl. Phys. Lett. 44, 898( 1984).

'>S. Datta, J. K. Furdyna, and R. L. Gunshor, Superlattices and Microstructures I, 327 (1985).

43R. R. Galazka, in Physicso fSemiconductors 1978, edited by B. 1. H. Wil-son (Institute of Physics Conf. Ser. No. 43, London, 1979), p. 133.

44J. K. Furdyna, J. Appl. Phys. 53, 7637 (1982).45N. B. Brandt and V. V. Moshchalkov, Adv. Phys. 33, 193 (1984).46A. V. Nurmikko, R. L. Gundshor, and L. A. Kolodziejski, J. Quantum

Electron (to be published).47M. Dobrowolska, A. M. Witowski, J. K. Furdyna, T. Ichiguchi, H. D.

Drew, and P. A. Wolff, Phys. Rev. B 29,6652 (1984).4KJ. A. Gaj. R. R. Galazka, and M. Nawrochi, Solid State Commun. 25, 193(1978).

49J. K. Furdyna and 1. Kossut, Superlattices and Microstructures (to bepublished) .

P. Kowalczyk, J. T. Cheung, E. A. Kraut, an d R. W. Grant, in Proceedings of The 1985 US Workshop on Physics and Chemistry of CdHgTe,San Diego, Oct. 7-10, 1985, p. 143 (unpublished).

SIL. A. Kolodziejski, T. Sakamoto, R. 1. Gunshor, and S. Datta, Appl .Pbys. Lett. 44, 799 (1984).

52J. 1. Nishizawa, private communicatio n and a lecture in The nnd National Symposium of American Vacuum Society, Houston, TX, Nov. 19-22,1985 (unpublished).

s ~ . M. Bedair, M. A. Tischler. T. Katsuy ama, an d N. A. EI-Masry, Appl.Phys. Lett. 47, 51 (1985).

:lAM. Tammenmaa, T. Koskinen, L. Hiltunen, M. Leskelii., and L. NiinistO,C. H. L. Goodman and M. V. Pessa R80

Downloaded 29 Nov 2011 to 210.212.129.125. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions


18/18

in Extended Abstracts ofThe First Symposium on Atomic Layer Epitaxy,Espoo, Finland, Dec. 13-14, 1984 (unpublished)."Y . P. Tanninen, M. Oikkonen, and T. Tuomi, Phys. Status Solidi A 67,573 (1981).

S6y. P. Tanninen and T. Tuomi, Thin Solid Films 90,339 (1982)."R . Tornqvist, J. Appl. Phys. 54,4110 (1983).SHy. P. Tanninen, M. Oikkonen, and T. Tuomi, Thin Solid Films 109,283(1983).s ~ . Tammenmaa (private communication).tooo. Thesis, H. Oppolzer, G. Ebbinghaus, and S. Schild, J. eryst. Growth63,47 (1983).

61J. A. Lahtinen and T. Tuomi, Acta Poly ech. Scand. Appl. Phys. Ser. 138,

97 (1983).62R. O. Tornqvist, J. Antson. J. Skarp. and V. P. Tanninen, IEEE TranElectron Devices ED-30, 468 (1983).63M. Oikkonen, M. Blomberg, T. Tuomi, and M. Tammenmaa. Thin SolFilms 124. 317 (1985).64L. E. Tannas, Jr., Electroluminescent Displays, Report 6475 (AerojElectro Systems. 1983).6SJ. A. Lahtinen, A. Lu, T. Tuomi, and M. Tammenmaa, J. Appl. Phys. 51851 (1985).

'"'T. Suntola, International Symposium Digest, Society of Information Diplay, New York, April 20-21, 1981 (unpublished).67T. Suntola and I. Hyvarinen, Annu. Rev. Mater. Sci. 15,177 (1985).

japplphys 60 alp

Documents