Surface & Coatings Technology 201 (2007) 9055 9059 www.elsevier.

com/locate/surfcoat

Thermal stability, sublimation pressures and diffusion coefficients of some metal acetylacetonates
M. Aslam Siddiqi, Rehan A. Siddiqui, Burak Atakan
Thermodynamics, IVG, Faculty of Engineering, University of Duisburg Essen, Lotharstr. 1, 47048 Duisburg, Germany Available online 21 April 2007

Abstract Metal acetylacetonates are often used as precursors for the deposition of oxides in CVD or ALE experiments. The thermal stability and the vapour/sublimation pressures of such compounds are of fundamental importance for reproducible depositions. We have started a program to study the long term thermal stability, sublimation pressures and the diffusion coefficients of organometallic compounds. The latter are needed for the calculation of the Sherwood and Lewis numbers used to describe mass transfer process. In a recent publication [M. A. Siddiqi, B. Atakan, Thermochim. Acta 452 (2007) 128.] we demonstrated a method for the evaluation of diffusion coefficients of less volatile substances in gas mixtures by combining the experimental thermogravimetric measurements with accurate sublimation pressure data (e.g. from Knudsen effusion method). In the present communication the results for aluminium(III) acetylacetonate, chromium(III) acetylacetonate, iron(III) acetylacetonate, manganese(III) acetylacetonate and nickel(II) acetylacetonate are presented. The last three of these show decomposition at low temperatures. Aluminium(III) acetylacetonate and chromium(III) acetylacetonate evaporate without residuals and hence their sublimation pressures and binary diffusion coefficients at various temperatures are reported for the first time. 2007 Elsevier B.V. All rights reserved.
Keywords: Sublimation pressure; Diffusion coefficients; Chromium acetylacetonate; Aluminium acetylacetonate; Iron acetylacetonate

1. Introduction For chemical vapour deposition (CVD) process the precursor molecules (which often are metalorganic compounds) are evaporated. After the evaporation one or more precursor molecules that include the elements which shall be present in a deposited thin film (or coating), are mixed and flown to a substrate. There thermal energy is provided to initiate a chemical reaction so that films of metals, oxides or other compounds are formed. To engineer such a process, the knowledge of the vapour or sublimation pressures and the stability at vaporization temperatures is essential because they determine the maximum theoretical growth rate and the composition. Thermal instability at these intermediate temperatures (typically 40150 C) may lead to pyrolysis in the evaporator and hence to low and non-reproducible growth rates in the CVD process. A number of metal acetylacetonates are used as CVD precursors, which are often easily commercially available and not too difficult to handle.

Corresponding author. Tel.: +49 203 3793355; fax: +49 203 3791594. E-mail address: b.atakan@uni-due.de (B. Atakan). 0257-8972/$ - see front matter 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2007.04.036

We have started a program to study the long term thermal stability, sublimation pressures and the diffusion coefficients of organometallic compounds. Recently the thermal behaviour of cobalt diketonate complexes and ferrocene [1,2] was reported. In the present communication the evaporation behaviour of aluminium acetylacetonate [Al(acac)3], chromium acetylacetonate [Cr(acac)3], iron acetylacetonate [Fe(acac)3], manganese acetylacetonate [Mn(acac)3] and nickel acetylacetonate [Ni(acac)2] is reported. An investigation of the stability of [ Fe(acac)3], [Mn(acac)3] and [Ni(acac)2] is very much needed as we find a number of publications reporting the sublimation/vapour pressures of these over a wide temperature range [39] although the decomposition of these has been reported in other publications in the past [1013]. [Al(acac)3] and [Cr(acac)3] evaporate without leaving any residue. Precise vapour pressure measurements for these substances are still required as the values of vapour pressures reported in literature [3,7,8,1418] differ from each other (details are given in the Results and discussion section). Furthermore, the diffusion coefficients are needed for the calculation of the Sherwood and Lewis numbers used to describe mass transfer processes (e.g. to calculate the buffer gas saturation in evaporators). The diffusion coefficient data for such substances

9056

M.A. Siddiqi et al. / Surface & Coatings Technology 201 (2007) 90559059

is scarce in literature and no data is available for the above mentioned substances. The diffusion coefficients have been obtained by the method developed earlier combining thermogravimetric measurements and the precise vapour pressures obtained from the Knudsen cell method. The main idea behind this method is that the evaporation out of a crucible is a nearly one dimensional diffusion process, which depends on the vapour pressure and the diffusion coefficient; if either is known, the other can be obtained from thermogravimetric mass loss rates. 2. Experimental Al(acac)3 (Sigma Aldrich, N 99%), Cr(acac)3 (Merck, N 99%), Fe(acac)3 (Merck, N 99%), Mn(acac)3 (Alfa aesar, N 99%) and Ni(acac)2 (Merck, N 98%) were purchased. For vapour pressure measurements these were purified by sublimation. Some measurements were also performed using untreated samples. The vapour pressures of the purified samples did not differ from those of the untreated samples. Hence for further thermogravimetric experiments the substances were used as such. N2 (99.98%) or He (99.998%) was used as a carrier gas. A home built stainless steel Knudsen cell was used for vapour pressure measurements. The details of the experimental set up and the method are given in a previous publication [1]. The Knudsen cell is situated in a stainless steel vessel, a vacuum chamber, with good thermal contact around the cell. The temperature of the stainless steel chamber was controlled with a PID temperature controller. The heating was done with an electrical band heater which was wrapped around the chamber carefully to cover it completely. The outer side was then covered with insulation material. The temperature was measured at two different places inside the chamber and did not differ more than 0.1 K. The difference between the actual evaporation temperature (inside the Knudsen cell) and the measured chamber temperature was determined in many experiments performed before the actual measurements by bringing a calibrated Pt-100 thermometer inside the Knudsen cell and measuring its temperature. This temperature difference which was different for different temperatures was always taken into account to correct the display temperature. It was ensured that the thermal equilibrium between the sample and the thermal reservoir (chamber) was attained. The Knudsen cell was evacuated with the help of a vacuum system consisting of a turbo molecular pump (Pfeifer TMH 071P), a fore pump (diaphragm backing pump, Pfeifer MVP 055-3) and measured by a pressure gauge (Pfeifer TPG 261). The pressure in the system was always below 10 7 Pa during each experimental run. A well defined amount of the substance (depending upon the temperature of the measurement and the substance) was weighed (accuracy: 0.03 mg) into the cell. The filling and weighing was done under inert gas atmosphere. The cell was then tightened and put into the vacuum chamber. The temperature of the chamber was maintained constant to better than 0.2 K. Prior to evacuation, enough time (at least 30 min) was allowed for the attainment of a constant temperature which was recorded with the help of a calibrated Pt-100 thermometer. After evacuating the vacuum chamber the time was measured between the time when the vacuum reached the pressure of

around 10 3 Pa and the time when the high vacuum pump was turned off and the pressure was above 10 3 Pa. Typical times were 13 h (in this time the weight losses were between 20 and 50 mg depending on the hole size, the temperature and the substance). The cell was then brought to room temperature in a desiccator and weighed again. The system was regularly tested with reference substances (ferrocene [1], phenanthrene [1], anthracene and pyrene) having different vapour pressures and proved to furnish reliable results over a large temperature range. More details are given in [1]. The uncertainties in the evaporation time and in the mass loss are estimated to be 0.5 min and 0.05 mg, respectively. In the evaluation of the data, no additional calibration was performed. The maximum overall uncertainty in vapour pressure measurements was estimated to be 0.1 to 0.5 Pa in the pressure range 1050 Pa and 0.02 to 0.1 Pa in the pressure range 0.410 Pa. This overall uncertainty was calculated assuming that the uncertainties in evaporation time, mass loss and the correction factor are independent and random. A commercial TGA/DTA (Bhr STA 503) was used to perform the thermogravimetric experiments. The apparatus as well as the micro balance was kept in a glove box to avoid any contact of the sample with oxygen or water vapour. The carrier gas was N2 or helium. The flow rate of 100 cm3/min (for nitrogen) was controlled by a calibrated mass flow controller. This flow rate was found to be sufficient to ensure that the concentration of substance at the top of the crucible remains nearly zero throughout the measurement as proved experimentally: a change in flow rate did not change the mass loss rate measurably. The pressure was throughout atmospheric. Open alumina crucibles were used in all experiments; the inner diameter being 5.35 mm, the inner height is 7.2 mm. The samples were filled inside the crucible, so that the initial height of the sample inside the crucible was between 3.0 and 5.0 mm. From the initial height, which was measured in the beginning of each experiment with an estimated accuracy of 0.2 mm, the volume of the sample was calculated. Using the initial mass the apparent density of the sample was derived. The temperature was typically reached within 30 min, and then the temperature was held constant, in some experiments until all the sample was evaporated, in other experiments the temperature was changed after 2 h to the next temperature, so that several temperatures could be investigated within one run. The temperature sensor was calibrated by measuring the melting points of reference substances (4-nitrotoluene, naphthalene, indium and potassium perchlorate) which cover the whole temperature range for the measurements. The uncertainty in mass loss rate was estimated to be 1% and 10% for the apparent density and for the diffusion coefficient 0.005 to 0.015 cm2 s 1 depending on the system and arising mainly from the uncertainties in the initial distance between the sample surface and the top of the crucible. This overall uncertainty in diffusion coefficient was calculated considering the uncertainties in vapour pressure, initial height of the sample and the mass loss rate given above to be independent and random. 3. Results and discussion The stability of the compounds was investigated by isothermal thermogravimetry (TGA). These experiments try to

M.A. Siddiqi et al. / Surface & Coatings Technology 201 (2007) 90559059

9057

368.4 K, 387.8 K and 407.3 K, respectively. Hoene et al. [12] reported its decomposition around 443 K and Beech and Lintonbon [10] reported 41% residual mass in the temperature range 452 K535 K. A typical evaporation curve, at 387.8 K is shown in Fig. 1b. For Ni(acac)2 vapour pressure data exist for temperatures up to 602 K [1618]. We find a residual of 76.6% even at 329.8 K (see in Fig. 1b). The temperature rise used to obtain the results of Fig. 1a was 1 C/min. For Fe(acac)3, Mn (acac)3 and Ni(acac)2 decomposition accompanies the evaporation process even at lower temperatures therefore no vapour pressure measurements could be obtained with the Knudsen cell method, which, as a gravimetric method, relies on the intact evaporation of the studied compound. For Al(acac)3 and Cr(acac)3 the vapour pressures were measured at various temperatures. The vapour pressure results for Al(acac)3 and Cr(acac)3 are shown in Fig. 2 as a function of 1/T; some literature values are also included. The enthalpy of sublimation derived from these vapour pressure values (from the slopes of the ln(p) vs. 1/T plots) is 101.75 kJ/mol and 128.17 kJ/mol respectively for Al(acac)3, and Cr(acac)3.

Fig. 1. TGA curves in non-isothermal mode (a) and isothermal mode (b) for the acetylacetonates. 1: Ni(acac)2, 2: Mn(acac)3, 3: Fe(acac)3, 4: Al(acac)3, 5: Cr(acac)3.

simulate typical evaporator conditions, which are held at constant temperatures. From theory in isothermal TGA a nearly linear mass loss as a function of time is expected, and no residual should remain. Some typical results of conventional and isothermal TGA are summarized in Fig. 1. A number of vapour pressure studies for Fe(acac)3 are found in the literature at temperatures up to 420 K [39]. On the other hand Beech and Lintonbon [10] found about 8.8% residual mass even at temperatures below 535 K. Varnek et al. [13] analysed the nature of volatility of iron compounds from Mssbauer data and reported only about 90% mass loss for Fe(acac)3. At constant temperatures of 407.35 K and 427.15 K, residual masses of 32.3% and 25.5%, respectively, were observed, which indicates evaporation was accompanied by the decomposition. This is shown in Fig. 1b for one temperature, 427.15 K. One can also see from the curves for Ni(acac)2, and Mn(acac)3, that even in the beginning the mass loss is not linear which indicates that the process is not exclusive evaporation of the precursor. For Mn(acac)3 we found residual masses of 37.8%, 47.8% and 66.3% constant temperatures of

Fig. 2. Vapour pressure of Cr(acac)3 (a) and Al(acac)3 (b).

9058

M.A. Siddiqi et al. / Surface & Coatings Technology 201 (2007) 90559059

The vapour pressure as a function of temperature was fitted to an Antoine expression, with pressures in kPa and temperatures in K: log p=kPa Ai Bi =T =K Ci : 1

The Antoine equation constants are given in Table 1. Our values for Al(acac)3 agree within experimental error with those of Sachindis et al. [8] and Malkerova et al. [15] but differ from those of Teghil et al. [18]. In [18] torsion effusion method has been used whose accuracy depends on the reliable values for the temperature and the geometrical factors which are difficult to check. Also the slope of ln(p) vs. 1/T curve is quite different and hence the derived enthalpy of sublimation. The enthalpy of sublimation derived from [18] is much lower than the other reported literature values. Lazarev et al. [16] report vapour pressures at higher temperatures. The vapour pressures at these higher temperatures calculated with our Antoine equation constants given in Table 1 agree with the reported values [8,15,16]. For Cr(acac)3 our values lie nearer to those reported by Malkerova et al. [3] and Semyannikov et al. [14]. Lazarev [16] values also agree with ours except at lower temperatures. The vapour pressure values reported by Pankajavalli et al. [17] are much lower than ours and those of Alikhanyan and Malkerova [15] are much higher. Pankajavalli et al. [17] use transpiration method in which the basic requirement is the achievement of equilibrium condition which may lead to systematic errors. Regarding the discrepancy in [15] it may be noted that the values reported by the same authors in a later publication [3] agree with our values. The values reported by Melia et al. [19] also differ slightly from ours. In summary a more consistent picture of the vapour pressures is emanating from the present

Fig. 3. Diffusion coefficient of Al(acac)3 at various temperatures: mean value (nitrogen), mean value (helium), temperature dependence T2, best fit (T 3.65 for helium and T 3.45 for nitrogen).

data sets. Also the good quality of the regularly performed vapour pressure measurement of reference substances lets us believe that the present data should be very reliable. The thermogravimetric experiments with Al(acac)3 and Cr (acac)3 were performed at five isotherms between 350 K and 480 K in nitrogen (or helium) gas atmosphere. The product of vapour pressure and the binary diffusion coefficient of the substances in carrier gas was calculated from the experimental mass loss (m) for the time t using the equation: n o RT q A pvap DAB h H 2 H 2 A 2Mt ( ) 2 Dm RT qA : 2 H H 2 SqA 2Mt Here T is the temperature, R the universal gas constant, A the apparent density of the evaporating substance of molecular mass M. H is the initial distance between the surface (of surface area S) of the material and the top of the crucible in thermogravimetric experiments. The experimental vapour pressure determined from the Knudsen effusion method was then used to derive the diffusion coefficient. Figs. 3 and 4 show the diffusion coefficients of Al(acac)3 and Cr(acac)3 in nitrogen and in helium. The maximum, minimum and the mean values for at least three runs are shown. As

Table 1 The constants for Eq. (1) Substance Ai Bi Ci subHm(exp) (temperature) / kJmol 1 101.75 (345410 K) subHm (l) (temperature) / kJmol 1 111 4 a (388413 K) 105 2 a (388413 K) 47 1 b (298 K) 102 3.2 c (432464 K) 127.28 0.22 d (320478 K) 113.0 4.8 c (357486 K) 111.6 3.0 e (374418 K) 110.77 0.2 f (320478 K)

Al(acac)3

13.55

6126.8

0.0

Cr(acac)3

13.21

5999.9

0.0

128.17 (345410 K)

a b c d e f

Ref. [8]. Ref. [18]. Ref. [16]. Ref. [14]. Ref. [17]. Ref. [19].

Fig. 4. Diffusion coefficient of Cr(acac)3 at various temperatures: mean value (nitrogen), mean value (helium), temperature dependence T2, best fit (T1.03 for helium and T0.82 for nitrogen).

M.A. Siddiqi et al. / Surface & Coatings Technology 201 (2007) 90559059

9059

expected the binary diffusion coefficients in helium are throughout higher than in nitrogen; also some lower diffusion coefficients for the chromium acetylacetonate compared to the one for aluminium acetylacetonate were expected due to the differences in their molecular mass. The difference between the values of the diffusion coefficients of Cr(acac)3 and Al(acac)3 is so large, that some dimerization of the chromium compound cannot be excluded. Also, the temperature dependence differs strongly for both compounds. From simple gas kinetic theory a Tn dependence with n = 1.752 is expected [20], while values of n near 3.5 (3.46 for N2, 3.66 for He) were found for the aluminium compound and values around n = 1 (0.82 for N2, 1.03 for He) were found for the chromium compound. These best fits were included together with a curve for n = 2 to the diagrams. Further theoretical investigations of these findings would be needed to gain a deeper understanding of this behaviour. The correlation for diffusion coefficient is given by D12 cm2 s1 2:38 1010 T 3:46 for Alacac3 in nitrogen D12 cm2 s1 1:45 1010 T 3:66 for Alacac3 in helium D12 cm2 s1 4:45 104 T 0:82 for Cracac3 in nitrogen D12 cm2 s1 2:93 104 T 1:03 for Cracac3 in helium : 4. Summary The thermal stability of five metal acetylacetonates, which are often used as CVD precursors, was investigated. It turned out that three of these (Fe, Mn, Ni) start to decompose at typical evaporator conditions, while the acetylacetonates of chromium and aluminium are stable at such temperatures. The vapour pressures were measured for these two compounds, resolving some inconsistent information from the literature. For the first time diffusion coefficients for these two stable diketonates were measured, giving the possibility of calculating the behaviour of CVD evaporators more easily.

Acknowledgement The authors like to thank Mr. A. Grnt and Mr. M. Richter for technical support. Financial support of this work by the DFG as part of the SPP 1119 Inorganic materials from the gas phase is gratefully acknowledged. References
[1] M.A. Siddiqi, B. Atakan, Thermochim. Acta 452 (2007) 128. [2] B. Atakan, M.A. Siddiqi, Proceedings Fifteenth European Conference on Chemical Vapor Deposition, 9, The Electrochemical Society Inc., U. S. A., 2005, p. 229. [3] I.P. Malkerova, A.S. Alikhanyan, V.G. Sevastyanov, Y. Kh. Grinberg, V.I. Gorgoraki, J. Inorg. Chem., USSR 35 (1990) 413 (Engl.Transl.). [4] J.L. Wood, M.M. Jones, Inorg. Chem. 3 (1964) 1533. [5] B.D. Fahlman, A.R. Barron, Adv. Mater. Opt. Electron. 10 (2000) 223. [6] E.W. Berg, J.T. Truemper, Anal. Chim. Acta 32 (1965) 245. [7] T.P. Melia, R. Merrifield, J. Inorg. Nucl. Chem. 32 (1970) 2573. [8] J. Sachindis, J.O. Hill, Thermochim. Acta 35 (1980) 59. [9] P.P. Semyannikov, I.K. Igumenov, S.V. Trubin, I.P. Asanov, J. Phys. France IV, 11 (2001) 995. [10] G. Beech, R.M. Lintonbon, Thermochim. Acta 3 (1971) 97. [11] R.G. Charles, M.A. Pawlikowski, J. Phys. Chem. 62 (1958) 440. [12] J.N. Hoene, R.G. Charles, W.M. Hickman, J. Phys. Chem. 62 (1958) 1098. [13] V.A. Varnek, I.K. Igumenov, L.N. Mazalov, J. Struct. Chem. 42 (2001) 860. [14] P.P. Semyannikov, I.K. Igumenov, S.V. Trubin, T.P. Chusova, Z.I. Semenova, Thermochim. Acta 432 (2005) 91. [15] A.S. Alikhanyan, I.P. Malkerova, Ya. Kh. Grinberg, V. B Lazarev, V.A. Bogdanov, V.I. Gorgoraki, V.A. Schreides, Russ. J. Phys. Chem. 292 (1987) 376 (Engl.Transl.). [16] V. B Lazarev, J.H. Greenberg, Z.P. Ozerova, G.A. Sharpataya, J. Therm. Anal. 33 (1988) 797. [17] R. Pankajavalli, C. Mallika, O.M. Shreedharan, P. AntonyPremkumar, K.S. Nagaraja, V.S. Rangunath, Chem. Eng. Sci. 57 (2002) 3603. [18] R. Teghil, D. Ferro, M. Pelino, Thermochim. Acta 44 (1981) 213. [19] T.P. Melia, R. Merrifield, J. Inorg. Nucl. Chem. 32 (1970) 1489. [20] B.E. Poling, J.M. Prausnitz, J.P. OConnell, The Properties of Gases and Liquids, fifth ed, McGraw-Hill, New York, 2001.