J. MATER. CHEM., 1994,4(8), 1239-1244 In situ Study of a Strontium P-Diketonate Precursor for Thinlfilm Growth by Atomic Layer Epitaxy Jaan Aarik,*" Aleks Aidla," Andres Jaek," Markku Leskela*b and Lauri Niinisto*" a Laboratory of Electroluminescence and Semiconductors, Tartu University, EE-2400 Tartu, Estonia b Department of Chemistry, University of Helsinki, FIN-0001 4 Helsinki, Finland Laboratory of Inorganic and Analytical Chemistry, Helsinki University of Technology, FIN-02 150 Espoo, Finland Precursor properties of Sr(thd), (thd =2,2,6,6-tetramethyl heptane-3,5-dione) have been investigated using in situ mass monitoring of thin films during atomic layer epitaxy growth cycles. H20 and H2S were used as the other source materials. The Sr(thd), source temperature, T,, significantly affects the growth rate.At T, =240-270 "C, a decrease of growth rate and in some cases even etching of the grown film takes place. These phenomena can be explained by the decomposition of Sr(thd),, which also explains the dependence of the growth rate on the reactor temperature, T,, observed when T,<240 "C and 240 <T,/"C <400. The growth experiments were complemented and the conclusions supported by separately studying (by mass spectrometry thermogravimetry, and differential thermal analysis) the thermal stability and fractionation of the precursor. The interest in strontium compounds has increased recently owing to the fact that SrS doped with rare-earths ions is a possible matrix/activator candidate for full-colour, thin-film electroluminescent (TFEL) displays.' In the development of materials for full-colour TFEL devices the blue phosphor has been the most difficult to achieve.For this purpose, SrS :Ce3+ has extensively been studied with considerable success.2 For instance, a full-colour EL device recently demonstrated was based on SrS : Ce and ZnS :Mn emitting layer^.^ SrO, on the other hand, is a constituent of high-T, superconducting oxides4y5 and of dielectrics for high-capacitance-density thin-film capacitors.6 For these applications a controllable depos- ition of SrS and SrO thin films is of great importance. In general, the alkaline-earth metals form very few volatile complexes which can be applied in thin-film deposition from the gas phase. The P-diketonates, especially the Sr(thd), complex (thd =2,2,6,6-tetramethyl heptane-3,5-dione) are among the few sufficiently stable volatile complexes of stron- tium.The P-diketonate polyether complexes form another group of volatile alkaline-earth-metal compounds recently reported in the Sr( thd), has been successfully applied in the SrS deposition by the atomic layer epitaxy (ALE) method.' ALE is a novel deposition technique in which the reactants are pulsed separately onto the substrate where exchange reactions take place and a monolayer of the desired product (or fraction thereof) is formed;" for instance, in the case of SrS the overall reaction is Sr(thd),(g) +H,S(g)+SrS(s) +2 Hthd(g). The actual reaction mechanism is more complicated owing to the thermal instability of the precursor Sr( thd),.11,12 Indeed, it has been shown that M(thd), as well as M(thd), M,(thd), (M =Ca, Sr) and other dissociation fragments can be found in the gas phase when thd chelates of alkaline-earth metals are ~olatilized.'~*'~Oligomeric and mixed (0x0 or aqua) ligand complexes can easily form with Ba p-dike ton ate^.'^,^^ In order to obtain a better understanding of the reaction mechanism and thus improved control of the deposition process and the resulting thin film properties, in situ studies are highly desirable.It was recently shown that monitoring the film mass during an ALE cycle is a powerful method which yields valuable information concerning adsorption and exchange reaction^.'^ In this study we have employed this method for in situ investigation of the ALE growth of SrO and SrS, where Sr(thd), and H,O or H2S were used as source materials. Ex situ thermogravimetry (TG), differential thermal analysis (DTA) and mass spectrometry (MS) studies of the precursor were also carried out. Parallel to the present study on Sr(thd),, we have also investigated the precursor properties of Ca(thd),18 and Ba(thd),." Experimental Experiments were carried out in a hot-wall flow-type ALE reactor (Fig.1). Fused quartz was used for all components of the reactor which were in contact with active gases at high temperatures. Argon was used as the carrier gas. Its pressure measured at the reactor outlet was kept at 150-200 Pa while the flow rate was 4m s-l inside the reactor tube.The commutation of the active gases was performed uia electro-magnetic valves operated by a personal computer. A gas manifold enabled us to switch the reactant flows on and off within 0.5-1.0s.The reactor temperature, TR,as well as the temperature of the effusion cell used as the Sr(thd), source, T,, could be controlled with an accuracy of 0.3 K. During the experiments TR and Ts were varied in the ranges 200-420°C and 200-340 "C, respectively. The Sr(thd), source chemical was synthesized as described earlier,12 and its purity and volatility were routinely checked for each batch by X-ray diffraction (XRD), differential scan- ning calorimetry (DSC)and TG/DTA. H20 or H2S were used as the other source materials. More detailed investigations into the volatility and gas- phase fractionation were carried out by TG/DTA measure- reactant gas inlet reactor mass sensor qocouple thermocouple solid reactant evaporation cell Fig.1 Schematic diagram of the flow-type hot-wall ALE equipment used ments in vucuo and by MS. For the thermoanalytical studies a Seiko TG/DTA 320 instrument of the series SSC 5200 was used and operated at a pressure of cu. 700Pa. The typical heating rate/sample weight combination was 10"C min-' and 5 mg. The UHV high-resolution mass spectra were recorded in a JEOL MS DX-303/DA5000 instrument. The solid samples were introduced into the mass spectrometer through a sample stage equipped with programmable heating.The mass of the deposited films in the ALE reactor was determined by a quartz crystal mass sensor. The sensor had Ag electrodes and its resonance frequency was 30 MHz. The films were grown directly on the Ag electrodes or on a buffer layer of aluminium or tantalum oxide. The increments of oscillation period, AT, proportional to the mass changes were plotted with a time resolution of 1s. The proportionality constant between AT and the corresponding mass change was (1.12+0.04)x 106gcm-2 s-'. Before starting the measurements a layer of SrS or SrO was grown onto the mass sensor using 20-50 ALE cycles. An ALE cycle included the injection of the Sr thd complex and H20 or H2S into the reactor, the reactants being separated by an inert- gas pulse.The first purge time following the injection of the thd complex was chosen to be sufficiently long (25 s) in order to avoid intermixing of the source materials in the gas phase; the second purge, following H,O or H,S injection, was set even longer (50-200 s). This time was needed to obtain the necessary data for the correction of experimental curves in respect of the temperature drift. Indeed, as the temperature coefficient of the mass sensor was as high as 530fs K-' at 250 "C, temperature increments as low as 0.01 K had a notice- able effect on AT. Therefore the reaction heat, valving of the gas flows as well as the instability of the reactor heater could affect the mass sensor signal. The temperature drift caused by the last factor was a slow one and was taken into account using the extrapolation of the time dependences of AT recorded during the second purge time.To minimize the role of the other two factors, appropriate cycle times were used. As the time constant of the mass sensor response to the temperature changes was ca. 10s the mass increments AT',, corresponding to the complete ALE cycle were recorded no sooner than 50-100 s after the end of the cycle. During the AzfO/Azl ratio measurements, where AT^ is the mass sensor signal increment caused by the adsorption of 80 1 I I Sr(thd), on H20off 1 J. MATER. CHEM., 1994,VOL. 4 Sr-containing precursor, we used short (1---2 s) exposures to the Sr(thd), source to reduce the concurrent temperature efTec t s.Results The oscillation period (AT) of the quartz crystal mass sensor as a function of time at two source temperatures is shown in Fig. 2, which also shows the different sequence of the pulse and purge times (tl-t4). The figure shows that an increase of T, from 235 to 265 "C has a significant effect on the curves. At 235°C the behaviour of the oscillation period resembles that observed for AlC13/H20, Ca( thd),/H,O and Ca(thd),/H,S precursor sy~terns.'~-'~ Similarly, the increase of the mass sensor signal during the time interval tl can be interpreted as the adsorption of metal-containing precursor, while the decrease of AT during the time interval t3 is mostly caused by the replacement of the heavy ligands by lighter oxygen atoms or OH groups.The specific features of the present case are that: (i) during exposure to the Sr(thd), source, the mass sensor signal does not reach a constant level but slowly continues to increase with the exposure time and (ii) the value of AT changes during the purge times. Plotting AT as a function of time under conditions when there was no Sr(thd), present as source chemical enabled us to determine that the increase of AT during time interval t, is caused by the decrease of the sensor temperature following switching-off the H,O flow. The origin of the incomplete saturation of AT, observed during the Sr-containing precursor pulse, is evident from Fig. 3 which shows that the saturation of the mass sensor signal increment AT'^ corresponding to a complete ALE cycle is also incomplete.As the temperature effects were neglected during AZ'~measurements, we concluded that this incomplete saturation as well as the decrease of AT during the time interval t, were caused by continuous CVD- type reactions which could be due to the decomposition of the source material or irreversible exchange reactions between adsorbate and H20 traces in the carrier gas. However, refer- ence measurements with the AlCl, precursor show that the role of the H20 background is negligible. The adsorption of A1C13 was self-limiting and no change of AT was observed after closing the AlCl, valve in spite of the fact that AlCl, is more reactive towards H20 than Sr( thd),. Consequently, the 40 I H,O on 0 -40 \ ri H200ff -80 Sr(thd), offP -80 1 JI I 1 1 I I I I I 0 100 200 300 100 200 0L tls Fig.2 Increment of mass sensor oscillation period A7 at TR=250 "C as a function of time at two different source temperatures (a) 235 "C and (b)265 "C J. MATER. CHEM., 1994, VOL. 4 0 a 16 24 pulse duration/s Fig.3 Increment of mass sensor oscillation period for a complete ALE cycle Af0 as function of thd complex pulse duration at TR= 280 "C and T,=290 "C. AT'^ obtained by extrapolation of the exper- imental data is the increment of mass sensor signal caused by the pure ALE growth during one ALE cycle. thermal decomposition of the precursor seems to be the only reason for the steady-state oscillation period increase. Fig.3 also demonstrates that the rate of this process, (dz/dt),,,, (slope of the broken line in Fig. 3), is more than an order of magnitude lower than the value of the derivative d(Az',)/dt recorded at low tl values. The latter determines the lowest deposition rate which can be limited by the arrival of precursor molecules. Therefore the thermal decomposition rate, which at all reactor temperatures was significantly lower, could not be transport-limited. Plotting (dz/dt)fi,,l us. reciprocal reactor temperature (Fig. 4), we find that the activation energy of the decomposition is 39 kJ mol-'. In order to avoid the uncer- tainties caused by this unintentional process, the values of Az0 presented in Fig. 5 and 6 were obtained via linear extrapol- ation of the respective experimental data to t, =0 as shown in Fig. 3.In this way quantities corresponding to the 'pure' ALE were obtained. The most unexpected phenomenon observed in this study is that AT, as a function of cycle time, changes its shape at some combinations of TR and Ts [Fig. 2(b)]. Moreover, in some cases when Ts increases, etching occurs instead of growth [Fig. (b)].The etching of the as-grown film took place at TR=250-265 "C, while at other TR values a decrease of the growth rate in the Ts region 240-250°C was observed [Fig. S(u)]. Relevant information concerning this anomalous behaviour was obtained from the dependence of the derivative 40 30 20 10 0 I I I I-1 0 1 2 1 240 260 280 source T/"C 1.0-m 0.6 -r9 -0.4 C-c h-+-.F -s0.2 0.1 I I 1 I I I I 1 1.4 1.6 1.8 2.0 lo3 T~IK Fig. 4 Rate of the steady-state oscillation period increase (dt/dt)f,,,l as function of l/TR (dz/dt)initi,l on the source temperature. Here (dz/dt)i,iti,l is the slope of the mass sensor signal us. time curve at the beginning of the Sr(thd), pulse. At this moment the surface coverage of Sr(thd), is small compared with the saturation level and the adsorption rate, which is proportional to dz/dt, is also pro- portional to the partial pressure of active species and their sticking coefficient. Fig. 5(b) shows that (dz/dt)initi,l has a sharp maximum at those source temperatures where a decrease of growth rate occurs, provided that the data were recorded during the increase of T,.This maximum indicates an anomal- ous increase in the vapour pressure and/or the activity and is most plausibly connected with the decomposition of Sr( thd), in the effusion cell. This supposition is consistent with the fact that no maxima or minima were recorded during the decrease of source temperature. It is interesting to note that at source temperatures below 240"C, where no anomalies were observed, Az0 as a function of the reactor temperature has a minimum which is also located at 240-250 "C [Fig. 6(u)].A steep increase at lower TR is mostly due to the unsaturated physisorption of Sr precursor. The increase in AT^ at TR3250 "C is very similar to the increase in the growth rate observed earlier', and will be discussed below.Note that Azo is nearly constant at Ts>270 "C [Fig. 6(b)]and its value is close to that obtained at Ts<240 "C and TR=380-400 "C. Fig. 6(b) also demon- strates that there is no significant difference in AT^ when H20 is replaced by H,S. Evidence indicating the decomposition of the Srl thd), precursor enables us to expect that the Sr complex which is chemisorbed on the solid surface in the ALE process is not source T/"C Fig. 5 Increments of mass sensor oscillation period for a complete ALE cycle AT^ (a) and derivative of oscillation period (dz/dt)initial (h) as a function of thd complex source temperature. (a)@, TR=320 "C; 0,TR=250 OC; (b) TR=320 "C. J. MATER.CHEM., 1994, VOL. 4 I I 1 40 20 PU 200 280 360 440 200 280 360 440 reactor TIoC Fig. 6 Increment of mass sensor oscillation period for a complete ALE cycle Ar0 as a function of reactor temperature for (a) T, =230 "C and (b) Ts=280 "C. 0,Sr(thd),/H,S; 0,Sr(thd),/H,O. 100 200 300 400 500 TI'C Fig.7 TG and DTA curves showing the volatilization and partial decomposition of Sr(thd),. The sample weight is 16.8 mg and the heating rate is loomin-' at a pressure of 700 Pa. Sr(thd),. In order to obtain data concerning this adsorbate we measured the ratio AZ'~/AZ~,where Az~is the change of the mass sensor signal increment caused by the adsorption of the Sr-containing precursor. This ratio is equal to the corre- sponding mass ratio Amo/Am, and, provided that the product of the complete ALE cycle is known, makes it possible to estimate the metal-to-ligand ratio in the chemisorbed com-plex.17 At TR,T, 2270 "C we obtained Amo/Am, =0.44 f0.06 and 0.46 0.05 for H2S and H20, respectively.At TR =200 "C and Ts=235 "C, Am,/Aml =0.25 0.05 when H,S was used as source material. With H,O the values of Amo/Aml range from 0.3 to 0.6 at TR,Tsd 240 "C. Such a large dispersion is probably due to incomplete exchange reactions between adsorbed thd complexes of strontium and H20 at these low temperatures. The results described above can be compared with those obtained in ex situ TG and MS studies. Fig. 7 shows typical thermoanalytical curves for the Sr( thd), precursor. According to the TG and DTA curves the volatilization at 220-320°C is rapid and smooth, but a residue (ca.10%) and the peaks in the DTA curve indicate decomposition and probably also melting. The small weight loss just above lOO"C, seen in the TG and especially in the DTA curve, is caused by the release of moisture and low-molecular-weight decomposition prod- ucts and its amount varies from batch to batch. The weight residue depends somewhat on the experimental conditions (heating rate, sample size etc.) but is typically 5-10%. That low-molecular-weight fragments are expelled from the precursor is corroborated by the mass chromatograms recorded at source temperatures of 100, 200 and 300°C (Fig. 8). Typical mass spectra corresponding to these plateaus are presented in Fig.8. The dominant peaks at m/z 54, 127, 271 and 725 correspond to C3H20+, C3H202C(CH3)3f, I 1 I I I I I I I 20 40 60 80 100 120 140 300oc scan no. l04III39 57 95 111 scan no. Fig. 8 Mass chromatogram of Sr(thd), showing the relative intensities of the main peaks as function of temperature (see the heating profile) Sr(thd) and Sr,(thd),, respectively. The peak at mjz 391, most notably present during the fast heating from 100 to 200"C, is probably a ligand-recombination product as it was found also in the spectra of Ca(thd), and Ba(thd),.,' The occurrence of Sr(thd) and Sr,(thd), at higher temperatures is in agreement with the behaviour of the corresponding calcium precur- Note, however, that the volatilization temperatures in various experiments (TG, ALE, MS) are not directly comparable because of the pressure differences.Discussion Our experimental results demonstrate that Sr( thd), as a source material for ALE growth displays noticeable peculiari- ties in the temperature range 240-270°C. As the behaviour of the Sr(thd), complex is not reversible it can be concluded that Sr( thd), partly decomposes at these temperatures. This conclusion is in agreement with the data published by Schwarberg et ~1.'~reporting that Sr( thd), melts at 248-268 "C with decomposition and also with the present thermoanalyt- ical and MS studies, which show that the gas phase already contains a significant amount of decomposition products (Fig. 9). Using the assumption of decomposition, a satisfactory explanation to all our results can be given.The explanation is based on the fact that the free ligand or its frag-ments volatilize at these temperatures more easily than an Sr-containing complex. One can also compare this case with J. MATER. CHEM., 1994, VOL. 4 'O01 127 271 I I 'II 100 200 300 400 500 600 700 800 mlz Fig. 9 Mass spectra of Sr(thd), obtained at (a) lOO"C, (b) during heating from 100 to 200 "C and (c)at 200 "C Ce(thd),, for which it has been shown that mass spectra of the vapour phase contain a number of peaks corresponding to dissociation fragments which do not contain metal atoms.21 Interacting with the surface of Sr-containing films, these thd fragments form an adsorbate that restricts the adsorption of the strontium complex.Furthermore, the as-formed adsorbate can be desorbed, resulting in etching.22 However, as the etching was not observed at reactor temperatures exceeding 265 "C, the process should be limited by the surface reactions rather than by the desorption rate. The experimental results enable us to conclude that the reaction between the free ligand and solid surface is exother- mic. Indeed, the time dependence of AT characterizing etching [Fig. 2(b)]shows that the sensor temperature decreases when the Sr source is switched off. This behaviour can only be caused by a preceeding increase of temperature because there was no such decrease in the cases when the growth was regular [Fig.2(u)] and when there was no Sr(thd), in the crucible. Consequently, the reaction rate between the solid film and the free ligand should decrease with increasing reactor tem- perature. This supposition is in good agreement with our experimental data where etching of the grown film took place in a narrow range of reactor temperatures above 250 "C while only a decrease of growth rate was observed at higher TR. The dependence of growth rate on the reactor temperature observed at TR 3240 "C and T,<240 "C [Fig. 6(u)] can also be explained by the decomposition of Sr(thd),. In this case, Sr(thd), decomposes on the substrate or in the reactor tube before reaching the substrate. All the dissociation fragments interact with the solid surface simultaneously and their contri- bution depends only on the reactor temperature.As the reaction between the solid film and the free ligand is exother- mic the effect of the latter should be reduced at higher temperatures. Thus, the growth rate should have a minimum at the lowest reactor temperatures where Sr( thd), decomposes. One can see from Fig. 5(u) that this conclusion is in good agreement with experimental data. Unfortunately the existing experimental data do not enable us to ascertain which kind of ligand-dissociation fragments are responsible for reducing the growth rate and etching. Correspondingly, it cannot be concluded whether or not these reactive intermediates can be neutralized, e.g.by incorporating additional components into the precursor.However, by pre- heating the precursor at Ts>270"C the effect of these dis- sociation products on the growth rate can be decreased. Assuming that at T,3270 "C Sr(thd), is, to a considerable degree, decomposed and that the free ligands are completely removed from the source material, the blocking of substrate surface by free ligands should be absent at these source temperatures. Consequently, the growth rate cannot depend on the reactor temperature in this case but is constant over a 1243 wide temperature range. As one can see from Fig. 5(h),our experimental data also confirm this conclusion. Additional information about less volatile decomposition products could be obtained from the chemical analysis of the residues which remain in the crucible and on the walls of the inlet tube.However, in our case the results of this analysis are not sufficiently reliable because we could not perform the analysis without exposing these reactive residues to ail-. As shown in our previous study,17 surface reactions can be determined using the ratio of the film mass increments corre- sponding to different steps of the ALE cycle. Unfortunately, the mass increments caused by the adsorption of the thd complex contain some experimental uncertainty which enables us to deduce only an approximate scheme of surface reactions. Assuming that thd ligands coordinated to Sr do not decom- pose during chemisorption, and excluding etching, the surface reactions for the first step of the ALE cycle can be expressed as Sr(thd),( g)+Sr(thd),(ads) +(x-y) thd(g) In some cases Sr( thd), can replace chemisorbed hydrogen: (x-y)H(ads)+Sr(thd),(g)+Sr(thd),(ads) +(x-y)H(thd)(g) (2) However, the values of Am, corresponding to these two cases differ by less than 1% and this difference cannot be used as basis for making any decision concerning the type of reaction.The reactions which take place during the second step of the ALE cycle can be expressed as Sr( thd),(ads) +H,S(g) SrSH, -,(ads) +yH (thd) ( g) (3) provided that the bonds between Sr and solid surface are stable. Similar equations can be written when H2S is replaced by H20. However, in the latter case the analysis is more compli- cated because Sr(OH), can be formed instead of SrO.Moreover, we still do not have an authentic corroboration of exchange reactions between thd complex of strontium and H20 being complete. Therefore the values of y can only be estimated for reactions between Sr(thd), and H,S. This was done using the experimental values of Amo/Am,. We estab- lished that at 270 <(TR,TS)/"C<400 the y value calculated according to eqn. (3) is 1.03f0.24. Hence at temperatures above 270°C Sr(thd), has lost a thd ligand before and/or during adsorption. Our MS results show that in the gas phase y is mainly 1.5. Thus the thd complex decomposes further on the substrate surface. This conclusion is also confirxned by the data presented in Fig. 3 and 4, which show that the deposition does not stop while the substrate is exposed to the Sr source and the role of this unsaturated adsorption increases with the increasing substrate temperature.For comparison only, we also calculated the value of y for reactions where H20 takes part instead of H2S. The values range from 0.6 to 1.0 at 270<(TR,Ts)/"C<400 and from 0.3 to 1.7 at 200 <(TR, Ts)/"C <240. We also estimated the mass of ligand(s) coordinated to one Sr atom adsorbed during the first step of the ALE cycle at TR=200 "C and Ts=235 "C. The m/z of this ligand calculated from Amo/Am, is 390 f100. This value is the mass of d single thd ligand multiplied by a factor of 2.1f0.6 and, within the limits of experimental error, coincides with the mass of a ligand coordinated to an Sr atom in the correspondrng MS measurements.Thus one can conclude that during adsorption at such a low temperature Sr-containing thd complexes do not lose any ligands. It is obvious that the size of a thd ligand is greater than that of an Sr site. Accordingly thd ligands included in the adsorbed particles limit the growth rate. In this connection 1244 J. MATER. CHEM., 1994, VOL. 4 the increase of the number of ligands per Sr atom with decreasing temperature explains the corresponding decrease of growth rate very well. 3 Workshop, ed. V. P. Singh and J. S. McClure, Cinco Puntos Press, 1992. M. Leppanen, G. Harkonen, A. Pakkala. E. Soininen and R. Tornqvist, Eurodisplay 93, Conference Proceedings, Conclusions This study demonstrates that Sr(thd), behaves in a compli- cated way as a source material for ALE growth.At source temperatures below 240°C the growth is rather stable. However, the growth rate depends on the reactor temperature. The molar mass of the main Sr-containing precursor adsorbed 4 5 Strasbourg, 1993, p. 229. T. Kimura, H. Nakao, H. Yamawaki, M. Ihara and M. Ozeki, ZEEE Trans. Magn., 1991,27,1211. J. M. Zhang, B. W. Wessels, D. S. Richeson, T. J. Marks, D. C. DeGroot and C. R. Kannevurf, J. Appl. Phys., 1991, 60, 2143. T. Sukuma, S. Yamamichi, S. Matsubara, H. Yamaguchi and Y. Miyasaka, Appl. Phys. Lett., 1990,57, 2431. J. A. T. Norman and C. P. Pez, J. Chem. Soc., Chem. Commun., appears to correspond closely to Sr(thd),. At the same time mass spectra indicate that some decomposition of thd ligands can take place during evaporation of the source material.At source temperatures ranging from 240 to 270°C the Sr source fractionates further. The vapour pressure of free ligands exceeds that of Sr-containing complexes. The adsorption of free ligands or their fragments dominates, resulting in very 10 1991,971. K. Timmer, K. I. M. A. Spee, A. Mackor. H. A. Meinema, A. L. Spek and P. van der Sluis, Znorg. Chim. Actu, 1991,190, 109. M. Tammenmaa, M. Asplund, H. Antson, L. Hiltunen, M. Leskela, L. Niinisto and E. Ristolainen, J. Crystal Growth, 1987,84, 151. M. Leskela and L. Niinisto, in Atomic Layer Epitaxy, ed. T. Suntola and M. Simpson, Blackie, Glasgow. 1990, p. 1. low or even negative values of growth rates (etching). At source temperatures above 270 "C, those dissociation fragments that do not contain Sr have mostly left the source material and the growth becomes stable again.Moreover, the main strontium complex is dimeric, giving rise to an increased 11 12 13 M. Leskela, Acta Polytech. Scand., Chem. Trrhnol. Metall. Ser., 1990, 195,67. M. Leskela, L. Niinisto, E. Nykanen, P. Soininen and M. Titta, Muter. Res. SOC.Symp. Ser., 1991,222, 315. J. E. Schwarberg, R. E. Sievers and R. W. Moshier, Anal. Chem., 1972,42,1828. growth rate, independent of the reactor temperature. 14 M. Leskela, L. Niinisto, E. Nykanen, P. Soininen and M. Tiitta, Thermochim. Acta, 1991,75,91. The authors thank A-A. Kiisler, K. Kukli and J. Laine-Ylijoki for experimental assistance. The cooperation of the Laboratory of Organic Chemistry at the Helsinki University of Technology (Prof.M. Lounasmaa and Mr. P. Sarkio) in obtaining the mass spectra is gratefully acknowledged. This 15 16 17 18 S. B. Turnipseed, R. M. Barkley and R. E. Sievers, Inorg. Chem., 1991,30,1164. J. G.Hubert-Pfalzgraf, Appl. Organornet. Chem., 1992,6,627. J. Aarik, A. Aidla, A. Jaek, A-A. Kiisler and A-A. Tammik, Acta Polytech. Scand., Chem. Technol. Metall. Ser., 1990, 195,201. J. Aarik, A. Aidla, A. Jaek, M. Leskela and L. Niinisto, Appl. Surf. work was partly supported by HUMAL Electronics Ltd. (Tartu, Estonia) and by the Technology Development Centre (TEKES in Helsinki, Finland; grant 4173/91). 19 20 Sci., 1994,75, 33. J. Aarik, A. Aidla, A. Jaek, M. Leskela and L. Niinisto, to be published. M. Lounasmaa, P. Sarkio, M. Leskela and L. Niinisto, to be published. References 21 M. Leskela, R. Sillanpaa, L. Niinisto and M. Tiitta, Acta Chem. Scand., 1991,45, 1006. 1 2 M. Leskela and L. Niinisto, Muter. Chem. Phys., 1991,31,7. See e.g. Proceedings of Electroluminescence Workshops, Acta 22 F. Rousseau, A. Jain, T. T. Kodas, M. Hampden-Smith, J. D. Farr and R. Muenchausen, J. Muter. Chem., 1992,2,893. Polytechn. Scand., Ser. Appl. Phys., Ph 179, ed. M. Leskela and E. Nykanen, 1990 and Electroluminescence, Proc. 6th Znt. Paper 3/07129D; Received 2nd December, 1993