- Email: [email protected]
Effect of deposition temperature on the properties of CeO2 films grown by atomic layer deposition
Thin Solid Films 519 (2011) 4192–4195
Contents lists available at ScienceDirect
Thin Solid Films j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / t s f
Effect of deposition temperature on the properties of CeO2 films grown by atomic layer deposition P.J. King a,⁎, M. Werner a, P.R. Chalker a, A.C. Jones b, H.C. Aspinall b, J. Basca b, J.S. Wrench b, K. Black b, H.O. Davies c, P.N. Heys c a b c
Department of Materials, University of Liverpool, George Holt building, L69 3BX, UK Department of Chemistry, University of Liverpool, Crown Street, L69 7ZD, UK SAFC Hitech Power Road, Bromborough, Wirral, Merseyside, CH62 3QF, UK
a r t i c l e
i n f o
Article history: Received 12 February 2010 Received in revised form 31 January 2011 Accepted 2 February 2011 Available online 18 February 2011 Keywords: Cerium oxide High dielectric constant materials Thin films Atomic layer deposition X-ray diffraction Raman spectroscopy
a b s t r a c t Cerium oxide dielectric thin films have been grown on n-type silicon by atomic layer deposition using a monomeric homoleptic CeIV alkoxide precursor with water vapour. Herein we report the dielectric properties of CeO2 films deposited from tetrakis(1-methoxy-2-methyl-2-propanolate)cerium. The resulting films exhibit permittivities in the range 25–42 at 1 MHz with a strong dependency on the deposition temperature. The microstructural origin of this behaviour has been investigated. The as-deposited films were found to be crystalline and they exhibited the cubic fluorite structure for deposition temperatures in the range 150 °C to 350 °C. Variations in the crystallite sizes are governed by the deposition temperature and have been estimated using a Debye–Scherrer analysis of the X-ray diffraction patterns. The changing crystallite size correlates with changes seen in the triply-degenerate F2g first-order Raman line half-width at 465 cm− 1. It is concluded that the frequency dependency of the film dielectric properties is strongly influenced by the crystallite size which in turn is governed by the growth temperature. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Cerium oxide (CeO2) has attracted attention as a dielectric thin film material for potential applications in complementary metal oxide semiconductor (CMOS) and memory applications; both as a potential insulator on silicon [1,2] and germanium [3,4] and as a substitutional dopant in other gate materials systems [5]. CeO2 may also be useful for high refractive index films [6], in applications like optical waveguides (frequently in doped systems with additions like yttrium or gadolinium) [7]. The dielectric properties of cerium oxide make it a potential candidate for the replacement of conventional silicon oxynitride materials in CMOS technology to reduce power consumption. Researchers at the Toshiba Corporation [8] have previously investigated the deposition of CeO2 on Si(111) substrates via molecular beam epitaxy using Ce and O3 sources. At a deposition temperature of 650 °C they observed no silicon oxide at the interface between the substrate and high-κ film. Capacitance–voltage measurements made on this structure yielded a reported dielectric constant (κ) of ~50. For a marginally higher deposition temperature of 700 °C κ was observed to fall to ~ 10, which was attributed to the formation of a cerium silicate interlayer between the film and the substrate. CeO2 readily crystallises in the fluorite form but control over the particle
size formed is important due to the effect of grain boundary density on properties like ionic conductivity and dielectric response. As well as the thin film reports some bulk ceria studies have also been found to be relevant to this work especially those dealing with nanostructured materials [9] and the Raman spectroscopy of ceria [10–12]. This paper describes the dielectric properties of CeO2 thin films deposited by atomic layer deposition (ALD). ALD has become an enabling manufacturing process for the deposition of gate and memory dielectric thin films due to its excellent control over nanometre-scale deposition rates, low pin-hole density and step coverage. ALD has been employed previously to deposit CeO2 films using Ce(thd)4 and Ce(thd)3phen (thd — 2,2,6,6-tetramethyl-3,5heptanedionate, phen — 1,10-phenanthroline) precursors with O3 as a source of oxygen [13]. The focus of that study was to develop CeO2 as a buffer layer for high-temperature superconductors and ferroelectric Pb(Zr,Ti)O3 films deposited on silicon or sapphire. We have recently reported [14] the development of a CeIV alkoxide precursor, Ce (mmp)4, (mmp — 1-methoxy-2-methyl-2-propanolate) which can be used with water vapour to deposit CeO2 films. The as-deposited films are polycrystalline and the dependency of the dielectric properties on crystallite grain size is discussed. 2. Experimental procedure
⁎ Corresponding author. E-mail address: [email protected] (P.J. King). 0040-6090/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2011.02.025
Deposition of the CeO2 films was performed by liquid injection ALD using a modified Aixtron AIX 200FE AVD reactor fitted with a liquid
P.J. King et al. / Thin Solid Films 519 (2011) 4192–4195
injector system [15]. The delivery of cerium precursor was achieved using a 0.05 M solution of [Ce(mmp)4] (SAFC Hitech Ltd) in toluene. Deionised water was used as a source of oxygen. Details of the ALD growth conditions are shown in Table 1. ALD runs were performed at substrate temperatures of 150 °C, 200 °C, 250 °C, 300 °C and 350 °C. The CeO2 films were deposited on n-Si(100) wafers. X-ray diffraction (XRD) patterns were measured using a Rigaku Miniflex diffractometer using Cu Kα radiation (0.154051 nm, 40 kV, 50 mA) spanning a 2θ range of 20°–50° at a scan rate of 0.01°/min. Raman spectra were acquired using a Jobin-Yvon LabRam HR consisting of a confocal microscope coupled to a single grating spectrometer equipped with a notch filter and a CCD camera detector. The Raman measurements were taken using a confocal aperture of 300 μm to limit the light taken to the centre of the microscope lens and hence improve the quality of signal achieved. The measurements spanned 150–1500 cm− 1 with four accumulations and an exposure time of 30 s. All of the spectra were recorded in a backscattering geometry using an incident wavelength of 325 nm from a He–Cd laser. Cross sectional transmission electron microscopy (XTEM) was carried out on a JEOL 3010 to study the film thicknesses and the structure of the films. Specimens were constructed by fixing cleaved segments of CeO2 films face to face with a thin layer of epoxy and leaving them clamped on a heating stage to cure the adhesive. The specimen stacks were then attached to a grinding jig using superglue, ground perpendicular to the glue plane down to an 8 μm finish and then separated by soaking in acetone and the grinding was repeated on the reverse of the specimen. Once thinned to a thickness of ~50 μm copper slot grids were attached and the stacks were ion milled in a 10− 5 mbar vacuum, from both sides, at a cone angle of 4° to the grinding plane at a beam voltage of 5 keV until the instant a hole appeared at the glue interface. After brief milling at a lower voltage/angle to effectively clean the surrounding area the specimens were transported to the microscope which used an operating voltage of 300 kV during the analysis. To determine the electrical characteristics of the films capacitance–voltage (C–V), capacitance–frequency (C–f) and current–voltage (I–V) measurements were carried out using an Agilent E4980A precision LCR meter. Gold contacts with an area of 4.5 × 10− 4 cm2 were deposited on the films to form metal oxide semiconductor (MOS) capacitors. Aluminum was deposited onto the backside of the silicon substrate to form an ohmic back-contact. 3. Results and discussion Fig. 1 shows the XRD diffraction patterns recorded for films deposited at 150 °C, 250 °C and 350 °C. There is an arbitrary offset in the intensity axis between data sets and background features have been removed for clarity. At all temperatures the films have diffraction peaks centred about 28.6°, 33.1° and 47.5° corresponding to the cubic (111), (200) and (220) diffraction peaks respectively (ICDS card [43–1002]). The peaks show a broadening at lower deposition temperatures; most noticeably in the (111) peak. Diffraction scans with a slower scan speed were carried out in the region of the (111) peak to obtain full width at half-maximum (FWHM) data due to the (111) being the most distinct diffraction peak. In XRD peak broadening occurs due to effectively sampling
4193
Fig. 1. XRD measurements for three deposition temperatures; three diffraction features from the cubic structure present are highlighted.
fewer planes in a film of smaller average crystallite size and hence increasing the angular range of lesser intensity around the diffraction peak. Other factors that can have an impact on peak width are instrumental broadening effects and lattice strain contributions. Instrumental broadening is not a concern due to the small size of the grains resulting in peaks that are considerably broader than the system resolution. Strain effects are discussed later. The Debye–Scherrer formula [16] dg =
0:9λ β cosθ
ð1Þ
was used to estimate the average crystallite diameters, dg, from the (111) X-ray diffraction features for each of the specimens. Here 0.9 is the Scherrer shape factor, λ the X-ray wavelength, β the FWHM of the sample peak and θ is the scattering angle at the centre of the peak being measured. The measurements performed have the grain size changing from ~ 6 nm for the 150 °C sample, to ~23 nm for the 350 °C sample. It is worth noting that strain can also produce a broadening of the XRD diffraction peak. The FWHM values used may contain a contribution from a strain component, introducing some level of uncertainty in the absolute values obtained. Nevertheless there is a clear trend towards larger crystallite grain sizes at higher temperatures; this is most likely due to increased mobility of the deposited material with increased deposition temperature. Fig. 2 shows the ultraviolet-excited Raman spectrum of CeO2 films deposited at the same three substrate temperatures. A mode at ~520 cm− 1 is associated with the silicon substrate and has been
Table 1 Growth conditions used to deposit CeO2 films. Growth parameter
Value
Substrate temperature Evaporator temperature Reactor pressure Solvent Concentration Argon carrier gas flow ([Ce(mmp)4]/purge/H2O/purge) Number of growth cycles
150–350 °C on n-Si(100) 100 °C 1 mbar toluene 0.05 M 100 cm3 min− 1 (2/2/0.5/3.5 s) 300
Fig. 2. UV Raman measurements for three deposition temperatures; the removed peak asterisked is the 520 cm− 1 silicon mode. The F2g and A1g modes are highlighted.
4194
P.J. King et al. / Thin Solid Films 519 (2011) 4192–4195
removed for clarity (denoted with an asterisk). The measurements are otherwise unaltered except for an arbitrary shift on the intensity axis. The mode at ~465 cm− 1 is the first-order triply-degenerate F2g mode associated with the fluorite crystal structure — a symmetrical stretching of the CeO8 unit in the lattice. This mode confirms that the crystalline phase is cubic. The other mode seen is the second-order A1g symmetry mode. Seven such second-order modes are predicted but two have additional contributions from the Eg (~600 cm− 1 mode) and F2g (~1178 cm− 1 mode) symmetry elements [12]. The presence of the second-order modes is testament to the resonance Raman process which samples phonon contributions away from the centre of the Brillouin zone. A clear shift of the F2g mode to higher wavenumber values as well as a broadening of the band with decreasing temperature is seen. This broadening effect is due to reduced phonon lifetime with smaller grain size. The general peak shift is attributed to a softening of the chemical bonds for lower crystallite size at the lower deposition temperatures and an increase in lattice constant [17]. Below a grain size of 100 nm this mode is expected to show a width decrease with grain size following a cubic root dependency; above this size regime the mode's sensitivity to oxygen sub-lattice disorder effects can result in broadening. The Raman results correlate with the XRD data and show that the films contain cubic crystallites and that the grain size increases with increasing deposition temperature. It is difficult to attach an accurate grain size to these measurements due to other factors, such as lattice strain, that affect the Raman line width. However the general trend is the same as with the XRD results showing larger grain sizes as the growth temperature increases. Both are affected by strain in the same way. Cross section transmission electron microscopy (XTEM) was performed to elucidate the crystalline grain size effects evident in both the XRD and Raman measurements. Fig. 3 presents grain size measurements done by the XRD FWHM method and those measured from the XTEM micrographs. The insets included show the grain structure at each temperature. Due to the overlapping nature of the crystallites it is not feasible to give a definitive average grain size with XTEM either. Nevertheless it is clear that the trend of increasing grain size with increasing temperature is corroborated. Fig. 4 shows capacitance–voltage (C–V) plots from films grown at 150 °C, 250 °C and 350 °C with thicknesses of 51 nm, 50 nm and 44 nm respectively obtained by spectroscopic ellipsometry. Curves for test frequencies of 1 kHz and 1 MHz are plotted. The dielectric constant can be extracted from the capacitance values in strong accumulation taking into account the presence of a 1.6 nm thick native SiO2 layer with a dielectric constant of 3.9. For the
Fig. 3. Grain size measurements for three deposition temperatures via the XRD FWHM/ Debye Scherrer method and those taken from XTEM micrographs. Insets show examples of the grain structure at each temperature.
Fig. 4. Capacitance–voltage (C–V) characteristics of the CeO2 films deposited at three deposition temperatures.
sample grown at 150 °C the extracted κ value at 1 MHz is 42 and for the sample grown at 350 °C the value at 1 MHz is 25. These are relatively high values in comparison to amorphous HfO2 which typically has a κ of 18. At the lower measurement frequency the capacitance increases, most markedly for the sample grown at 150 °C. The change in capacitance with frequency is better illustrated in the C–f plot given in Fig. 5 which summarises the accumulation capacitance (1 V bias) as a function of bias frequency for the CeO2 films deposited at 150 °C and 350 °C. This plot shows significant frequency dispersion for the sample grown at 150 °C. There are five reasons which may cause the frequency dispersion observed: (i) series resistances, (ii) parasitic effects (including back contact imperfection and cables and connections), (iii) leakage currents, (iv) the interlayer between the high-κ layer and silicon substrate and (v) a κ value dependence on frequency of the CeO2 film. To obtain the genuine intrinsic properties and permittivity of the CeO2 dielectric from the CV measurements the first four effects must be eliminated. The effects of series resistances and parasitic effects were reported in
Fig. 5. Capacitance–frequency (C–f) characteristics of the CeO2 films deposited at 150 °C and 350 °C measured at a bias voltage of 1 V.
P.J. King et al. / Thin Solid Films 519 (2011) 4192–4195
4195
range 150 °C to 350 °C and hence the grain boundary network reduces. Grain boundaries are a well established source of leakage current pathways and consequently films with a smaller grain boundary surface area (e.g. larger average crystallite size) are expected to display lower leakage current densities in films of comparable thicknesses. In these samples the variation in leakage between samples is not very different but the leakage current density is relatively high in comparison with other crystalline hafnium oxide based dielectric materials measured under comparable conditions [5]. 4. Conclusions
Fig. 6. Leakage current densities for films at three deposition temperatures.
our previous work [18,19]. To minimise the effects of series resistances and back contact imperfections (including contact resistance R, contact capacitance C, or parasitic R–C coupled in series etc.) aluminium back contacts were deposited over a large area of the substrate wafer. The same procedure was carried out for all samples. All samples tested had the same or very similar substrate area (~2 × 2 cm2) to ensure that the effects of series resistance and back contact imperfections were the same for all samples. Furthermore measurement cables and connections were kept short to further minimize parasitic capacitance effects and were the same for all samples. Fig. 6 presents leakage current density data for the films at three different deposition temperatures. A measure of the leakage current density for films grown at the three temperatures of 150 °C, 250 °C and 350 °C is 9.4 × 10− 4, 6.6 × 10− 4 and 7.4 × 10− 4 A/cm2 respectively at a field of 0.5 MV/cm. The similarities in the data at voltages where the films are in strong accumulation (i.e. 0.4 to 0.6 MV/cm) between the different samples suggest that the changes in the C–V and C–f data are not due to leakage. In all cases the films were deposited on essentially identical Si wafers with a 1.6 nm native SiO2 layer and hence all have a similar interfacial layer. Deconvoluting these factors allows the material properties of the films to be ascertained. One potential explanation for the different frequency dispersion, or dielectric relaxation, behaviour observed between the different growth temperatures relates to the different crystal grain sizes formed at the different growth temperatures. It has been reported that a decrease in crystal grain size can cause an increase in the dielectric relaxation in ferroelectric relaxor ceramics; most likely due to increased stress within smaller crystal grains [20,21]. In addition to ferroelectric relaxor ceramics it has been reported that lanthanumdoped zirconium oxide high-κ dielectrics also suffer from a severe dielectric relaxation as the size of crystalline grains is reduced [18]. With these CeO2 films those consisting of ~6 nm nanocrystallites suffer a much more severe dielectric relaxation than those with ~ 23 nm nanocrystallites. We conclude that the physical process behind the relaxation is dominated by the size of the crystallite grains formed during growth. It is clear that films consisting of smaller crystallites will contain a larger grain boundary network than one consisting of larger crystallites. As the polarisability of a dielectric grain is mainly from the surface this could account for the large increase in capacitance observed in the 150 °C sample at 1 kHz which then reduces with frequency due to the dielectric relaxation effects. It is apparent that the crystallite size increases over the temperature
In conclusion ALD ceria thin films were found to deposit as a crystalline film for a range of substrate temperatures within the ALD growth window of the Ce[mmp]4 precursor with water as an oxidant. An X-ray diffraction Debye–Scherrer analysis shows an increase in crystallite size for higher growth temperatures. The first-order F2g Raman mode for ceria exhibits a shift to higher wavenumber and a peak broadening for smaller crystallites. A change in C–V behaviour and dielectric relaxation is observed for films grown at different deposition temperatures. The crystallite size has been observed to play a role in the dielectric relaxation behaviour. Acknowledgements PJK acknowledges the Engineering and Physical Sciences Research Council (EPSRC) for funding of his studentship. The authors are also grateful to SAFC Hitech for provision of the Ce[mmp]4 precursor. Thanks to K. Dawson for his assistance with the XTEM work. References [1] S. Logothetidis, P. Patsalas, E.K. Evangelou, N. Konofaos, I. Tsiaoussis, N. Frangis, Mater. Sci. Eng. B 109 (2004) 69. [2] S. Wang, W. Wang, J. Zuo, Y. Qian, Mater. Chem. Phys. 68 (2001) 246. [3] E.K. Evangelou, M.S. Rahman, A. Dimoulas, IEEE Trans. Electron. Devices 56 (2009) 399. [4] F.-C. Chiu, S.-Y. Chen, C.-H. Chen, H.-W. Chen, H.-S. Huang, H.-L. Hwang, Jpn. J. Appl. Phys. 48 (2009) 04C014. [5] P.R. Chalker, M. Werner, S. Romani, R.J. Potter, K. Black, H.C. Aspinall, A.C. Jones, C.Z. Zhao, S. Taylor, P.N. Heys, Appl. Phys. Lett. 93 (2008) 182911. [6] S. Kanakaraju, S. Mohan, A.K. Sood, Thin Solid Films 305 (1997) 191. [7] L. Bi, H.-S. Kim, G.F. Dionne, S.A. Speakman, D. Bono, C.A. Ross, J. Appl. Phys. 103 (2008) 07D138. [8] Y. Nishikawa, T. Yamaguchi, M. Yoshiki, H. Satake, N. Fukushima, Appl. Phys. Lett. 81 (2002) 4386. [9] S. Saitzek, J.-F. Blach, S. Villain, J.-R. Gavarri, Phys. Status Solidi (a) 205 (2008) 1534. [10] G. Gouadec, P. Colomban, Prog. Crys. Growth Charact. Mater. 53 (2007) 1. [11] A. Nakajima, A. Yoshihara, M. Ishigame, Phys. Rev. B 50 (1994) 13297. [12] W.H. Weber, K.C. Hass, J.R. McBride, Phys. Rev. B 48 (1993) 178. [13] J. Päiväsaari, M. Putkonen, L. Niinisto, J. Mater. Chem. 12 (2002) 1828. [14] J.S. Wrench, K. Black, H.C. Aspinall, A.C. Jones, J. Bacsa, P.R. Chalker, P.J. King, M. Werner, H.O. Davies, P.N. Heys, Chem. Vap. Deposition 15 (2009) 1. [15] R.J. Potter, P.R. Chalker, T.D. Manning, H.C. Aspinall, Y.F. Loo, A.C. Jones, L.M. Smith, G.W. Critchlow, M. Schumacher, Chem. Vap. Deposition 11 (2005) 159. [16] P. Scherrer, Gött. Nachr. 2 (1918) 98. [17] F. Zhang, S.-W. Chan, J.E. Spanier, E. Apak, Q. Jin, R.D. Robinson, I.P. Herman, Appl. Phys. Lett. 80 (2002) 127. [18] C.Z. Zhao, S. Taylor, M. Werner, P.R. Chalker, R.T. Murray, J.M. Gaskell, A.C. Jones, J. Appl. Phys. 105 (2009) 044102. [19] P. Taechakumput, C.Z. Zhao, S. Taylor, M. Werner, P.R. Chalker, J.M. Gaskell, A.C. Jones, M. Drobnis, in: Yang Ming (Ed.), Semiconductor Technology Conference (ISTC2008), Proceeding of the 7th International Conference on Semiconductor Technology, ISBN: 978-988-17408-1-6, 2008, p. 20. [20] H. Yu, H. Liu, H. Hao, L. Guo, C. Jin, Z. Yu, M. Cao, Appl. Phys. Lett. 91 (2007) 222911. [21] N. Sivakumar, A. Narayanasamy, C.N. Chinnasamy, B. Jeyadevan, J. Phys. Condens. Matter 19 (2007) 386201.
Contents lists available at ScienceDirect
Thin Solid Films j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / t s f
Effect of deposition temperature on the properties of CeO2 films grown by atomic layer deposition P.J. King a,⁎, M. Werner a, P.R. Chalker a, A.C. Jones b, H.C. Aspinall b, J. Basca b, J.S. Wrench b, K. Black b, H.O. Davies c, P.N. Heys c a b c
Department of Materials, University of Liverpool, George Holt building, L69 3BX, UK Department of Chemistry, University of Liverpool, Crown Street, L69 7ZD, UK SAFC Hitech Power Road, Bromborough, Wirral, Merseyside, CH62 3QF, UK
a r t i c l e
i n f o
Article history: Received 12 February 2010 Received in revised form 31 January 2011 Accepted 2 February 2011 Available online 18 February 2011 Keywords: Cerium oxide High dielectric constant materials Thin films Atomic layer deposition X-ray diffraction Raman spectroscopy
a b s t r a c t Cerium oxide dielectric thin films have been grown on n-type silicon by atomic layer deposition using a monomeric homoleptic CeIV alkoxide precursor with water vapour. Herein we report the dielectric properties of CeO2 films deposited from tetrakis(1-methoxy-2-methyl-2-propanolate)cerium. The resulting films exhibit permittivities in the range 25–42 at 1 MHz with a strong dependency on the deposition temperature. The microstructural origin of this behaviour has been investigated. The as-deposited films were found to be crystalline and they exhibited the cubic fluorite structure for deposition temperatures in the range 150 °C to 350 °C. Variations in the crystallite sizes are governed by the deposition temperature and have been estimated using a Debye–Scherrer analysis of the X-ray diffraction patterns. The changing crystallite size correlates with changes seen in the triply-degenerate F2g first-order Raman line half-width at 465 cm− 1. It is concluded that the frequency dependency of the film dielectric properties is strongly influenced by the crystallite size which in turn is governed by the growth temperature. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Cerium oxide (CeO2) has attracted attention as a dielectric thin film material for potential applications in complementary metal oxide semiconductor (CMOS) and memory applications; both as a potential insulator on silicon [1,2] and germanium [3,4] and as a substitutional dopant in other gate materials systems [5]. CeO2 may also be useful for high refractive index films [6], in applications like optical waveguides (frequently in doped systems with additions like yttrium or gadolinium) [7]. The dielectric properties of cerium oxide make it a potential candidate for the replacement of conventional silicon oxynitride materials in CMOS technology to reduce power consumption. Researchers at the Toshiba Corporation [8] have previously investigated the deposition of CeO2 on Si(111) substrates via molecular beam epitaxy using Ce and O3 sources. At a deposition temperature of 650 °C they observed no silicon oxide at the interface between the substrate and high-κ film. Capacitance–voltage measurements made on this structure yielded a reported dielectric constant (κ) of ~50. For a marginally higher deposition temperature of 700 °C κ was observed to fall to ~ 10, which was attributed to the formation of a cerium silicate interlayer between the film and the substrate. CeO2 readily crystallises in the fluorite form but control over the particle
size formed is important due to the effect of grain boundary density on properties like ionic conductivity and dielectric response. As well as the thin film reports some bulk ceria studies have also been found to be relevant to this work especially those dealing with nanostructured materials [9] and the Raman spectroscopy of ceria [10–12]. This paper describes the dielectric properties of CeO2 thin films deposited by atomic layer deposition (ALD). ALD has become an enabling manufacturing process for the deposition of gate and memory dielectric thin films due to its excellent control over nanometre-scale deposition rates, low pin-hole density and step coverage. ALD has been employed previously to deposit CeO2 films using Ce(thd)4 and Ce(thd)3phen (thd — 2,2,6,6-tetramethyl-3,5heptanedionate, phen — 1,10-phenanthroline) precursors with O3 as a source of oxygen [13]. The focus of that study was to develop CeO2 as a buffer layer for high-temperature superconductors and ferroelectric Pb(Zr,Ti)O3 films deposited on silicon or sapphire. We have recently reported [14] the development of a CeIV alkoxide precursor, Ce (mmp)4, (mmp — 1-methoxy-2-methyl-2-propanolate) which can be used with water vapour to deposit CeO2 films. The as-deposited films are polycrystalline and the dependency of the dielectric properties on crystallite grain size is discussed. 2. Experimental procedure
⁎ Corresponding author. E-mail address: [email protected] (P.J. King). 0040-6090/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2011.02.025
Deposition of the CeO2 films was performed by liquid injection ALD using a modified Aixtron AIX 200FE AVD reactor fitted with a liquid
P.J. King et al. / Thin Solid Films 519 (2011) 4192–4195
injector system [15]. The delivery of cerium precursor was achieved using a 0.05 M solution of [Ce(mmp)4] (SAFC Hitech Ltd) in toluene. Deionised water was used as a source of oxygen. Details of the ALD growth conditions are shown in Table 1. ALD runs were performed at substrate temperatures of 150 °C, 200 °C, 250 °C, 300 °C and 350 °C. The CeO2 films were deposited on n-Si(100) wafers. X-ray diffraction (XRD) patterns were measured using a Rigaku Miniflex diffractometer using Cu Kα radiation (0.154051 nm, 40 kV, 50 mA) spanning a 2θ range of 20°–50° at a scan rate of 0.01°/min. Raman spectra were acquired using a Jobin-Yvon LabRam HR consisting of a confocal microscope coupled to a single grating spectrometer equipped with a notch filter and a CCD camera detector. The Raman measurements were taken using a confocal aperture of 300 μm to limit the light taken to the centre of the microscope lens and hence improve the quality of signal achieved. The measurements spanned 150–1500 cm− 1 with four accumulations and an exposure time of 30 s. All of the spectra were recorded in a backscattering geometry using an incident wavelength of 325 nm from a He–Cd laser. Cross sectional transmission electron microscopy (XTEM) was carried out on a JEOL 3010 to study the film thicknesses and the structure of the films. Specimens were constructed by fixing cleaved segments of CeO2 films face to face with a thin layer of epoxy and leaving them clamped on a heating stage to cure the adhesive. The specimen stacks were then attached to a grinding jig using superglue, ground perpendicular to the glue plane down to an 8 μm finish and then separated by soaking in acetone and the grinding was repeated on the reverse of the specimen. Once thinned to a thickness of ~50 μm copper slot grids were attached and the stacks were ion milled in a 10− 5 mbar vacuum, from both sides, at a cone angle of 4° to the grinding plane at a beam voltage of 5 keV until the instant a hole appeared at the glue interface. After brief milling at a lower voltage/angle to effectively clean the surrounding area the specimens were transported to the microscope which used an operating voltage of 300 kV during the analysis. To determine the electrical characteristics of the films capacitance–voltage (C–V), capacitance–frequency (C–f) and current–voltage (I–V) measurements were carried out using an Agilent E4980A precision LCR meter. Gold contacts with an area of 4.5 × 10− 4 cm2 were deposited on the films to form metal oxide semiconductor (MOS) capacitors. Aluminum was deposited onto the backside of the silicon substrate to form an ohmic back-contact. 3. Results and discussion Fig. 1 shows the XRD diffraction patterns recorded for films deposited at 150 °C, 250 °C and 350 °C. There is an arbitrary offset in the intensity axis between data sets and background features have been removed for clarity. At all temperatures the films have diffraction peaks centred about 28.6°, 33.1° and 47.5° corresponding to the cubic (111), (200) and (220) diffraction peaks respectively (ICDS card [43–1002]). The peaks show a broadening at lower deposition temperatures; most noticeably in the (111) peak. Diffraction scans with a slower scan speed were carried out in the region of the (111) peak to obtain full width at half-maximum (FWHM) data due to the (111) being the most distinct diffraction peak. In XRD peak broadening occurs due to effectively sampling
4193
Fig. 1. XRD measurements for three deposition temperatures; three diffraction features from the cubic structure present are highlighted.
fewer planes in a film of smaller average crystallite size and hence increasing the angular range of lesser intensity around the diffraction peak. Other factors that can have an impact on peak width are instrumental broadening effects and lattice strain contributions. Instrumental broadening is not a concern due to the small size of the grains resulting in peaks that are considerably broader than the system resolution. Strain effects are discussed later. The Debye–Scherrer formula [16] dg =
0:9λ β cosθ
ð1Þ
was used to estimate the average crystallite diameters, dg, from the (111) X-ray diffraction features for each of the specimens. Here 0.9 is the Scherrer shape factor, λ the X-ray wavelength, β the FWHM of the sample peak and θ is the scattering angle at the centre of the peak being measured. The measurements performed have the grain size changing from ~ 6 nm for the 150 °C sample, to ~23 nm for the 350 °C sample. It is worth noting that strain can also produce a broadening of the XRD diffraction peak. The FWHM values used may contain a contribution from a strain component, introducing some level of uncertainty in the absolute values obtained. Nevertheless there is a clear trend towards larger crystallite grain sizes at higher temperatures; this is most likely due to increased mobility of the deposited material with increased deposition temperature. Fig. 2 shows the ultraviolet-excited Raman spectrum of CeO2 films deposited at the same three substrate temperatures. A mode at ~520 cm− 1 is associated with the silicon substrate and has been
Table 1 Growth conditions used to deposit CeO2 films. Growth parameter
Value
Substrate temperature Evaporator temperature Reactor pressure Solvent Concentration Argon carrier gas flow ([Ce(mmp)4]/purge/H2O/purge) Number of growth cycles
150–350 °C on n-Si(100) 100 °C 1 mbar toluene 0.05 M 100 cm3 min− 1 (2/2/0.5/3.5 s) 300
Fig. 2. UV Raman measurements for three deposition temperatures; the removed peak asterisked is the 520 cm− 1 silicon mode. The F2g and A1g modes are highlighted.
4194
P.J. King et al. / Thin Solid Films 519 (2011) 4192–4195
removed for clarity (denoted with an asterisk). The measurements are otherwise unaltered except for an arbitrary shift on the intensity axis. The mode at ~465 cm− 1 is the first-order triply-degenerate F2g mode associated with the fluorite crystal structure — a symmetrical stretching of the CeO8 unit in the lattice. This mode confirms that the crystalline phase is cubic. The other mode seen is the second-order A1g symmetry mode. Seven such second-order modes are predicted but two have additional contributions from the Eg (~600 cm− 1 mode) and F2g (~1178 cm− 1 mode) symmetry elements [12]. The presence of the second-order modes is testament to the resonance Raman process which samples phonon contributions away from the centre of the Brillouin zone. A clear shift of the F2g mode to higher wavenumber values as well as a broadening of the band with decreasing temperature is seen. This broadening effect is due to reduced phonon lifetime with smaller grain size. The general peak shift is attributed to a softening of the chemical bonds for lower crystallite size at the lower deposition temperatures and an increase in lattice constant [17]. Below a grain size of 100 nm this mode is expected to show a width decrease with grain size following a cubic root dependency; above this size regime the mode's sensitivity to oxygen sub-lattice disorder effects can result in broadening. The Raman results correlate with the XRD data and show that the films contain cubic crystallites and that the grain size increases with increasing deposition temperature. It is difficult to attach an accurate grain size to these measurements due to other factors, such as lattice strain, that affect the Raman line width. However the general trend is the same as with the XRD results showing larger grain sizes as the growth temperature increases. Both are affected by strain in the same way. Cross section transmission electron microscopy (XTEM) was performed to elucidate the crystalline grain size effects evident in both the XRD and Raman measurements. Fig. 3 presents grain size measurements done by the XRD FWHM method and those measured from the XTEM micrographs. The insets included show the grain structure at each temperature. Due to the overlapping nature of the crystallites it is not feasible to give a definitive average grain size with XTEM either. Nevertheless it is clear that the trend of increasing grain size with increasing temperature is corroborated. Fig. 4 shows capacitance–voltage (C–V) plots from films grown at 150 °C, 250 °C and 350 °C with thicknesses of 51 nm, 50 nm and 44 nm respectively obtained by spectroscopic ellipsometry. Curves for test frequencies of 1 kHz and 1 MHz are plotted. The dielectric constant can be extracted from the capacitance values in strong accumulation taking into account the presence of a 1.6 nm thick native SiO2 layer with a dielectric constant of 3.9. For the
Fig. 3. Grain size measurements for three deposition temperatures via the XRD FWHM/ Debye Scherrer method and those taken from XTEM micrographs. Insets show examples of the grain structure at each temperature.
Fig. 4. Capacitance–voltage (C–V) characteristics of the CeO2 films deposited at three deposition temperatures.
sample grown at 150 °C the extracted κ value at 1 MHz is 42 and for the sample grown at 350 °C the value at 1 MHz is 25. These are relatively high values in comparison to amorphous HfO2 which typically has a κ of 18. At the lower measurement frequency the capacitance increases, most markedly for the sample grown at 150 °C. The change in capacitance with frequency is better illustrated in the C–f plot given in Fig. 5 which summarises the accumulation capacitance (1 V bias) as a function of bias frequency for the CeO2 films deposited at 150 °C and 350 °C. This plot shows significant frequency dispersion for the sample grown at 150 °C. There are five reasons which may cause the frequency dispersion observed: (i) series resistances, (ii) parasitic effects (including back contact imperfection and cables and connections), (iii) leakage currents, (iv) the interlayer between the high-κ layer and silicon substrate and (v) a κ value dependence on frequency of the CeO2 film. To obtain the genuine intrinsic properties and permittivity of the CeO2 dielectric from the CV measurements the first four effects must be eliminated. The effects of series resistances and parasitic effects were reported in
Fig. 5. Capacitance–frequency (C–f) characteristics of the CeO2 films deposited at 150 °C and 350 °C measured at a bias voltage of 1 V.
P.J. King et al. / Thin Solid Films 519 (2011) 4192–4195
4195
range 150 °C to 350 °C and hence the grain boundary network reduces. Grain boundaries are a well established source of leakage current pathways and consequently films with a smaller grain boundary surface area (e.g. larger average crystallite size) are expected to display lower leakage current densities in films of comparable thicknesses. In these samples the variation in leakage between samples is not very different but the leakage current density is relatively high in comparison with other crystalline hafnium oxide based dielectric materials measured under comparable conditions [5]. 4. Conclusions
Fig. 6. Leakage current densities for films at three deposition temperatures.
our previous work [18,19]. To minimise the effects of series resistances and back contact imperfections (including contact resistance R, contact capacitance C, or parasitic R–C coupled in series etc.) aluminium back contacts were deposited over a large area of the substrate wafer. The same procedure was carried out for all samples. All samples tested had the same or very similar substrate area (~2 × 2 cm2) to ensure that the effects of series resistance and back contact imperfections were the same for all samples. Furthermore measurement cables and connections were kept short to further minimize parasitic capacitance effects and were the same for all samples. Fig. 6 presents leakage current density data for the films at three different deposition temperatures. A measure of the leakage current density for films grown at the three temperatures of 150 °C, 250 °C and 350 °C is 9.4 × 10− 4, 6.6 × 10− 4 and 7.4 × 10− 4 A/cm2 respectively at a field of 0.5 MV/cm. The similarities in the data at voltages where the films are in strong accumulation (i.e. 0.4 to 0.6 MV/cm) between the different samples suggest that the changes in the C–V and C–f data are not due to leakage. In all cases the films were deposited on essentially identical Si wafers with a 1.6 nm native SiO2 layer and hence all have a similar interfacial layer. Deconvoluting these factors allows the material properties of the films to be ascertained. One potential explanation for the different frequency dispersion, or dielectric relaxation, behaviour observed between the different growth temperatures relates to the different crystal grain sizes formed at the different growth temperatures. It has been reported that a decrease in crystal grain size can cause an increase in the dielectric relaxation in ferroelectric relaxor ceramics; most likely due to increased stress within smaller crystal grains [20,21]. In addition to ferroelectric relaxor ceramics it has been reported that lanthanumdoped zirconium oxide high-κ dielectrics also suffer from a severe dielectric relaxation as the size of crystalline grains is reduced [18]. With these CeO2 films those consisting of ~6 nm nanocrystallites suffer a much more severe dielectric relaxation than those with ~ 23 nm nanocrystallites. We conclude that the physical process behind the relaxation is dominated by the size of the crystallite grains formed during growth. It is clear that films consisting of smaller crystallites will contain a larger grain boundary network than one consisting of larger crystallites. As the polarisability of a dielectric grain is mainly from the surface this could account for the large increase in capacitance observed in the 150 °C sample at 1 kHz which then reduces with frequency due to the dielectric relaxation effects. It is apparent that the crystallite size increases over the temperature
In conclusion ALD ceria thin films were found to deposit as a crystalline film for a range of substrate temperatures within the ALD growth window of the Ce[mmp]4 precursor with water as an oxidant. An X-ray diffraction Debye–Scherrer analysis shows an increase in crystallite size for higher growth temperatures. The first-order F2g Raman mode for ceria exhibits a shift to higher wavenumber and a peak broadening for smaller crystallites. A change in C–V behaviour and dielectric relaxation is observed for films grown at different deposition temperatures. The crystallite size has been observed to play a role in the dielectric relaxation behaviour. Acknowledgements PJK acknowledges the Engineering and Physical Sciences Research Council (EPSRC) for funding of his studentship. The authors are also grateful to SAFC Hitech for provision of the Ce[mmp]4 precursor. Thanks to K. Dawson for his assistance with the XTEM work. References [1] S. Logothetidis, P. Patsalas, E.K. Evangelou, N. Konofaos, I. Tsiaoussis, N. Frangis, Mater. Sci. Eng. B 109 (2004) 69. [2] S. Wang, W. Wang, J. Zuo, Y. Qian, Mater. Chem. Phys. 68 (2001) 246. [3] E.K. Evangelou, M.S. Rahman, A. Dimoulas, IEEE Trans. Electron. Devices 56 (2009) 399. [4] F.-C. Chiu, S.-Y. Chen, C.-H. Chen, H.-W. Chen, H.-S. Huang, H.-L. Hwang, Jpn. J. Appl. Phys. 48 (2009) 04C014. [5] P.R. Chalker, M. Werner, S. Romani, R.J. Potter, K. Black, H.C. Aspinall, A.C. Jones, C.Z. Zhao, S. Taylor, P.N. Heys, Appl. Phys. Lett. 93 (2008) 182911. [6] S. Kanakaraju, S. Mohan, A.K. Sood, Thin Solid Films 305 (1997) 191. [7] L. Bi, H.-S. Kim, G.F. Dionne, S.A. Speakman, D. Bono, C.A. Ross, J. Appl. Phys. 103 (2008) 07D138. [8] Y. Nishikawa, T. Yamaguchi, M. Yoshiki, H. Satake, N. Fukushima, Appl. Phys. Lett. 81 (2002) 4386. [9] S. Saitzek, J.-F. Blach, S. Villain, J.-R. Gavarri, Phys. Status Solidi (a) 205 (2008) 1534. [10] G. Gouadec, P. Colomban, Prog. Crys. Growth Charact. Mater. 53 (2007) 1. [11] A. Nakajima, A. Yoshihara, M. Ishigame, Phys. Rev. B 50 (1994) 13297. [12] W.H. Weber, K.C. Hass, J.R. McBride, Phys. Rev. B 48 (1993) 178. [13] J. Päiväsaari, M. Putkonen, L. Niinisto, J. Mater. Chem. 12 (2002) 1828. [14] J.S. Wrench, K. Black, H.C. Aspinall, A.C. Jones, J. Bacsa, P.R. Chalker, P.J. King, M. Werner, H.O. Davies, P.N. Heys, Chem. Vap. Deposition 15 (2009) 1. [15] R.J. Potter, P.R. Chalker, T.D. Manning, H.C. Aspinall, Y.F. Loo, A.C. Jones, L.M. Smith, G.W. Critchlow, M. Schumacher, Chem. Vap. Deposition 11 (2005) 159. [16] P. Scherrer, Gött. Nachr. 2 (1918) 98. [17] F. Zhang, S.-W. Chan, J.E. Spanier, E. Apak, Q. Jin, R.D. Robinson, I.P. Herman, Appl. Phys. Lett. 80 (2002) 127. [18] C.Z. Zhao, S. Taylor, M. Werner, P.R. Chalker, R.T. Murray, J.M. Gaskell, A.C. Jones, J. Appl. Phys. 105 (2009) 044102. [19] P. Taechakumput, C.Z. Zhao, S. Taylor, M. Werner, P.R. Chalker, J.M. Gaskell, A.C. Jones, M. Drobnis, in: Yang Ming (Ed.), Semiconductor Technology Conference (ISTC2008), Proceeding of the 7th International Conference on Semiconductor Technology, ISBN: 978-988-17408-1-6, 2008, p. 20. [20] H. Yu, H. Liu, H. Hao, L. Guo, C. Jin, Z. Yu, M. Cao, Appl. Phys. Lett. 91 (2007) 222911. [21] N. Sivakumar, A. Narayanasamy, C.N. Chinnasamy, B. Jeyadevan, J. Phys. Condens. Matter 19 (2007) 386201.