- Email: [email protected]
Stability of MgO(1 1 1) films grown on 6H-SiC(0 0 0 1) by molecular beam epitaxy for two-step integration of functional oxides
Available online at www.sciencedirect.com
Applied Surface Science 254 (2008) 3191–3199 www.elsevier.com/locate/apsusc
Stability of MgO(1 1 1) films grown on 6H-SiC(0 0 0 1) by molecular beam epitaxy for two-step integration of functional oxides T.L. Goodrich, Z. Cai, K.S. Ziemer * Department of Chemical Engineering, Northeastern University, Boston, MA 02115, United States Received 25 July 2007; received in revised form 27 October 2007; accepted 30 October 2007 Available online 4 November 2007
Abstract ˚ thick, was grown by molecular beam epitaxy (MBE) on hydrogen cleaned hexagonal silicon Crystalline magnesium oxide (MgO) (1 1 1), 20 A carbide (6H-SiC). The films were further heated to 740 8C and 650 8C under different oxygen environments in order to simulate processing conditions for subsequent functional oxide growth. The purpose of this study was to determine the effectiveness and stability of crystalline MgO films and the MgO/6H-SiC interface for subsequent heteroepitaxial deposition of multi-component, functional oxides by MBE or pulsed laser deposition processes. The stability of the MgO films and the MgO/6H-SiC interface was found to be dependent on substrate temperature and the presence of atomic oxygen. The MgO films and the MgO/6H-SiC interface are stable at temperatures up to 740 8C at 1.0 109 Torr for extended periods of time. While at temperatures below 400 8C exposure to the presence of active oxygen for extended periods of time has negligible impact, exposure to the presence of active oxygen for more than 5 min at 650 8C will degrade the MgO/6H-SiC interface. Concurrent etching and interface breakdown mechanisms are hypothesized to explain the observed effects. Further, barium titanate was deposited by MBE on bare 6H-SiC(0 0 0 1) and MgO(1 1 1)/6H-SiC(0 0 0 1) in order to evaluate the effectiveness of the MgO as a heteroepitaxial template layer for perovskite ferroelectrics. # 2007 Elsevier B.V. All rights reserved. PACS : 81.15.Hi 79.60.i Keywords: MgO; SiC; Heteroepitaxy; Functional oxides; MBE; Wide bandgap semiconductor
1. Introduction The realization of next-generation devices for hightemperature, high-frequency applications will require the integration of multi-component, functional oxides on wide bandgap semiconductors, including hexagonal silicon carbide (6H-SiC) [1]. Multi-component, functional oxides, including multiferroics (bismuth ferrite), ferroelectrics (barium titanate), piezoelectrics (lead zirconate titanate), and ferrimagnetics (barium hexaferrite) are of interest because of the potential for controlled coupling between mechanical, electrical, and magnetic properties. In order for the successful integration of functional oxides on 6H-SiC, it is necessary to establish an abrupt and effective interface between the 6H-SiC and the functional oxide film. Chen et al. [2] reported on the improved crystallography and magnetic properties of M-type barium
* Corresponding author. Tel.: +1 617 373 2990; fax: +1 617 373 2209. E-mail address: [email protected] (K.S. Ziemer). 0169-4332/$ – see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2007.10.077
hexaferrite (BaM, BaFe12O19) grown by pulsed laser deposition (PLD) on 6H-SiC(0 0 0 1) through the use of a graduated magnesium oxide (MgO) interlayer. Meier et al. [3] demonstrated improved epitaxial growth of (0 0 1) oriented barium titanate (BTO) grown by metal organic chemical vapor deposition by integrating crystalline MgO(0 0 1) through molecular beam epitaxy (MBE) on epitaxial strontium titanate ˚ crystalline on Si(0 0 1). We have previously shown that a 20 A MgO(1 1 1) film deposited on 6H-SiC(0 0 0 1) by MBE provides an effective heteroepitaxial template and prevents interface breakdown during functional oxide film growth [4,5]. The bulk lattice spacing of MgO(1 1 1) suggests that the film acts as a buffer layer to improve lattice matching between the 6HSiC substrate and the desired functional oxide. For example, 6H˚ compared to the SiC(0 0 0 1) has a bulk lattice spacing of 3.08 A oxygen to oxygen lattice spacing of BaM(0 0 0 1), which is ˚ . The oxygen to oxygen lattice spacing of MgO(1 1 1) is 2.94 A ˚ , which when integrated as a heteroepitaxial template, can 2.98 A reduce lattice mismatch and aid in creating an effective interface through controlled heteroepitaxy. In order for the MgO layer to
3192
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
be an effective heteroepitaxial template, it is necessary to deposit smooth, single crystalline MgO(1 1 1) on the 6H-SiC(0 0 0 1) substrate. Several groups have demonstrated the ability to grow smooth crystalline MgO(1 1 1) or MgO(1 0 0) on the semiconductor substrates of gallium nitride [6,7], silicon [3], and silicon carbide [4,5,8] by MBE. Typical deposition conditions include temperatures ranging from room temperature up to 600 8C. For use of the MgO film as an interlayer between a ferroelectric, ferromagnetic, or multiferroic oxide, the substrate and MgO film may be exposed to different environmental conditions from the MgO film deposition conditions. In the MBE environment, this typically means temperatures up to 700 8C and active oxygen species from a plasma source or an ozone source [9–11]. Usually, however, the functional oxides are deposited through a physical deposition process, like PLD where the MgO/ SiC would be exposed to growth temperatures as high as 925 8C and even higher annealing temperatures, which can exceed 1000 8C, and millitorr or atmospheric pressures of oxygen [2]. Thus a variety of processing conditions, and perhaps two processing technologies (MBE and PLD), may be necessary to realize full integration of device quality functional oxides and semiconductors. This study takes the next step in realizing a twostep growth process by understanding the impact of different oxide processing environments on thin MgO films and the MgO(1 1 1)/SiC(0 0 0 1) interface. In this work, MgO thin films were deposited on 6HSiC(0 0 0 1) at a constant substrate temperature of 140 8C and an oxygen plasma at 5.0 106 Torr with a photomultiplier reading of 35 mV. These processing conditions result in an adsorption controlled growth regime where the growth rate of the MgO film increases linearly with respect to magnesium flux ˚ MgO films were previously found to be thermally [5]. The 20 A stable as indicated by X-ray photoelectron spectroscopy (XPS), reflection high-energy electron diffraction (RHEED), and atomic force microscopy (AFM), after heating to 740 8C at ˚ 1.0 109 Torr for 90 min. Similarly, heating 380 A MgO(1 1 1) films deposited on 6H-SiC(0 0 0 1) to 790 8C for 30 min in vacuum and in air were again thermally stable, as indicated by XPS, RHEED, AFM, and a Bruker D-5000 diffractometer with Highstar area detection [5]. Due to the ˚ MgO films, it was not possible to thickness of the 380 A characterize the interface by XPS, as the XPS sampling depth ˚ . Therefore, X-ray for Si and C through MgO is 93 A diffraction (XRD) was performed on the films. The XRD patterns contained diffraction peaks characteristic of only 6HSiC(0 0 0 1) and MgO(1 1 1), and did not indicate any additional phases or breakdown in the MgO/SiC after exposure. This paper reports the effects of post-deposition temperature and oxygen species exposure on the MgO(1 1 1) film and the MgO(1 1 1)/SiC(0 0 0 1) interface. Chemical, morphological, and crystallographic properties are compared before and after the exposures, which were chosen based on the typical processing requirements for subsequent MBE or PLD growth of multi-component functional oxides, but under extended time periods. The purpose of this extended time exposure was to understand the impact of a subsequent deposition environment on the MgO layers. As shown in this paper, certain
combinations of elevated temperature and oxygen species can detrimentally impact the film and interface. By understanding potential mechanisms of degradation, successful twostep growth processes can be developed in order to avoid conditions that result in interface breakdown during actual deposition of the multilayered heterostructure. 2. Experimental 6H-SiC(0 0 0 1) (Cree, Inc.) substrates were degreased in heated solvents of trichloroethylene, acetone, and methanol. The samples were then cleaned in a custom built hydrogen flow furnace in order to remove surface contamination and scratches resulting from mechanical polishing. The cleaned 6HSiC(0 0 0 1) surface consisted of a H3 H3 R30o silicate adlayer surface reconstruction, similar to the results reported by Bernhardt et al. [12]. Details and characterization of the clean 6H-SiC(0 0 0 1) surface are described elsewhere [4,5,13]. The samples were then mounted on a molybdenum puck with conductive silver paint, immediately loaded into a ultra-high vacuum system, and evacuated to a base pressure of 1.0 109 Torr. Chemical and bonding characterization was performed by XPS using a PHI model 04-173-0-007 Mg/Al dual anode, nonmonochromatic X-ray source and a PHI model 10-360-4-015 hemispherical analyzer. XPS peak deconvolution was performed using a minimum full width at half maximum of 1.6 eV and an 80/20 Gaussian/Lorenzian peak shape, as determined from a clean Au4f7 photoelectron peak. In situ surface crystallographic characterization was performed using a Staib Instruments RHEED system RH-15 and a k-Space Associates KSA 400 digital camera and software. Ex situ morphological characterization was performed using a NANOPROBE IIIA Digital Instruments AFM in tapping mode with an oxide sharpened SiN tip with a radius of curvature less than 10 nm. MgO was deposited by MBE using a SPECS Scientific Instruments dual source, low-temperature effusion cell filled with magnesium shavings (99.98%, Alpha Aesar) and an Oxford Applied Research remote oxygen rf-plasma source, model HD25. The plasma source was equipped with ion deflection plates located at the end of the discharge tube, which prevented any charged species from impinging on the surface during deposition. The absence of ions was confirmed by setting the ionization energy to zero on a Hiden Analytical HALO 201 quadrupole mass spectrometer, and turning the ion deflection plates of the plasma source on and off. The plasma source was also equipped with an optical photodiode filtered for a wavelength of 844 nm, which allowed direct evaluation of the atomic oxygen content in the plasma. Substrate temperature was measured using a two-color optical pyrometer and a type C thermocouple in contact with the backside of the molybdenum sample holder. In all experiments, the MgO films were deposited in a constant oxygen plasma environment with a background pressure of 5.0 106 Torr and a filtered photomultiplier reading of 35 mV. Although quantitative characterization of the oxygen plasma has not been performed, the optical photodiode
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
filter on the plasma discharge tube enables a qualitative measure of atomic oxygen species. An increase in atomic oxygen is directly associated with an increase in the photomultiplier reading. Through preliminary characterization, the maximum obtainable photomultiplier reading at this oxygen throughput was found to be around 400 mV. Comparing the photomultiplier reading of 35 mV used in this study with the maximum obtainable reading of 400 mV indicates that the atomic oxygen content of the oxygen plasma was relatively low. In all cases, the plasma source was operated with a voltage bias between the two deflection plates at the exit of the discharge tube, thus removing all oxygen ions in the plasma before they could reach the substrate surface. BTO films were deposited by MBE at substrate temperatures of 650 8C and constant oxygen plasma at 5.0 106 Torr with a filtered photomultiplier reading of 35 mV. The barium was supplied from a barium rod (99.9%, Electronic Space Products International) in the same SPECS Scientific Instruments dual source, low-temperature effusion cell as the magnesium. The dual source effusion cell consists of two independently controlled cells capable of supplying two different materials. In our system the source is set up with magnesium and barium. The titanium was supplied by a Varian Inc. mini Ti-Ball source, originally designed as a sublimation source for an ion pump, and is based off the design by Theis et al. [14]. Film thicknesses were calculated using the TPP-2M and Gries equations [15] and the XPS signal attenuation of the Si2p photoelectrons. These thickness calculations have an inherent error around 10% due to transport approximations, surface roughness, surface excitations, and surface refraction [15]. For the material parameters used in these equations, all thickness calculations were based on the physical properties of bulk, crystalline materials. It is fairly safe to assume bulk properties for all MgO film thickness calculations used in this study based on the findings by Li et al. [16], who reported strain relaxation of the crystal lattice to bulk properties for MgO(1 0 0) deposited on SrTiO3(1 0 0) at thickness of 1.75 monolayers ˚ . Secondary thickness measurements were carried out or 7.4 A ˚ MgO film, using transmission electron microscopy on a 25 A which confirmed the reliability of the attenuation thickness calculation within a 10% error. For some cases described in the results and discussion section below, a layered structure of MgO/SiOx/SiC is formed. In order to
3193
quantitatively analyze the formation of the SiOx layer, modified thickness calculations were necessary to take into consideration appropriate attenuation parameters due to the hypothesized MgO/SiOx/SiC layered structure. As described later, the assumption that the SiOx is located at the interface with negligible intermixing between the MgO and SiC was supported by XPS and RHEED, which indicated the crystalline MgO(1 1 1) pattern was maintained on the surface during all exposures. Since the MgO is assumed to be on the surface, the thickness of the MgO film after exposure was determined through a direct proportion between the area of the deconvoluted XPS Mg2p peaks and the calculated thicknesses of the as-deposited film. The overall attenuation of the silicon was known by comparing the XPS intensity before growth, after growth, and after exposure. Similarly, the attenuation of silicon through SiOx and MgO was determined as a function of varying thickness. Since the thickness of MgO after exposure was proportionally calculated, the thickness of the SiOx layer was iterated until the overall observed silicon attenuation was achieved. Note that the assumption of a layered structure increases the error associated with the thickness calculation. Although accurate thickness quantification is very difficult by this method, the associated error is consistent for each sample, which allows for qualitative comparison between the different exposure conditions. Thus arguments are made based on a relative comparison in order to support the qualitative analysis of undesirable interface degradation between the MgO and 6H-SiC. 3. Results and discussion ˚ of MgO(1 1 1) was In the first set of experiments, 20 A deposited on hydrogen cleaned 6H-SiC(0 0 0 1) at a substrate temperature of 140 8C, a magnesium flux of 1.0 1014 cm2 s1, an oxygen chamber pressure of 5.0 106 Torr, and a photomultiplier reading of 35 mV. These MgO films were then exposed to four different post-deposition conditions: (1) heated to 740 8C in vacuum (1.0 109 Torr) for 90 min, (2) heated to 650 8C in molecular oxygen (5.0 106 Torr) for 60 min, (3) heated to 650 8C in a light oxygen plasma (5.0 106 Torr, 35 mV) for 60 min, and (4) heated to 650 8C in a harsh plasma (5.0 106 Torr, 125 mV) for 60 min. The processing conditions for the as-deposited films and post-deposition exposure are listed in Table 1. RHEED was
Table 1 Sample identification with the processing parameters used Sample
Time (min)
Temperature (8C)
Mg flux (cm2 s1)
O2 pressure (Torr)
Photomultiplier (mV)
14
6
5.0 10 1.0 109
35 0
1
Pre Post
5 90
140 740
1.0 10 N/A
2
Pre Post
5 60
140 650
1.0 1014 N/A
5.0 106 5.0 106
35 0
3
Pre Post
5 60
140 650
1.0 1014 N/A
5.0 106 5.0 106
35 35
4
Pre Post
5 60
140 650
1.0 1014 N/A
5.0 106 5.0 106
35 125
Samples 1–4 represent the four as-deposited MgO films (pre) and post-deposition long-term exposures (post).
3194
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
used real-time to monitor the crystal orientation of the MgO films during growth and during post-deposition exposures. The four as-deposited MgO films (samples 1-pre, 2-pre, 3-pre, ˚. and 4-pre) were grown to an average thickness of 20.3 2.9 A ˚ The 2.9 A does not include the 10% error associated with the thickness calculations. Rather, it is indicative of the error associated with thickness variation between the four samples, ˚ . XPS which were grown to a target thickness of 20 A characterization of the four as-deposited MgO films indicated that they were stoichiometric with an average composition of MgO1.070.18. After each post-deposition exposure, XPS analysis indicated that the chemistry for all of the MgO films (samples 1-post, 2-post, 3-post, and 4-post) were stable with no change in bonding state of the Mg2p peak and no indication of magnesium peroxide, magnesium metal, or magnesium silicide. However, after exposure to oxygen at elevated temperature (samples 2-post, 3-post, and 4-post) the O1s and Si2p photoelectron peaks indicated the formation of a SiOx layer. Fig. 1 qualitatively illustrates the appearance of the SiOx layer as indicated by the Si2p XPS spectra. The peak intensities in Fig. 1 have not been normalized, and thus illustrate the additional signal attenuation of the silicon in bulk SiC through the SiOx layer. This additional attenuation was used to help calculate the thickness of the SiOx layer and follows the assumption that the SiOx layer is located at the MgO/SiC interface. RHEED indicated the crystalline MgO(1 1 1) pattern was maintained on the surface during all exposures. Thus, it is reasonable to assume that the SiOx is forming at the MgO/SiC interface and does not influence the crystal structure of the MgO film. The resulting MgO/SiOx/SiC layered structure is most evident in samples 3-post and 4-post. The SiOx layer is composed of two different oxidation states of the silicon. The left most peak, with a binding energy around 103.3 eV is associated with SiO2, while the middle peak, with a binding energy around 102.1 eV, is associated with SiO [17].
Fig. 1. XPS spectra of Si2p for samples 1–4 after heat/oxygen exposure illustrating the increased MgO/SiC interface breakdown with exposure to increased atomic oxygen. The SiOx formation for samples 3-post and 4-post consists of both SiO2 and SiO.
Sample 1, which was heated to 740 8C for 90 min at 1.0 109 Torr with no oxygen exposure, showed a slight ˚ . This change decrease in MgO film thickness of less than 0.2 A in thickness is within the error of the thickness calculation itself and is not considered significant. In addition, the attenuated signal originating from the silicate adlayer remained unchanged after exposure. This illustrates the thermal stability of the MgO/SiC interface when heated in vacuum. The MgO film of sample 2, which was heated to 650 8C under molecular oxygen at a pressure of 5.0 106 Torr for ˚ . This is a 15% decrease 60 min, decreased in thickness by 2.6 A in MgO thickness compared to the as-deposited film, which is outside the 10% calculation error and is a repeatable trend when ˚ MgO films in molecular oxygen. There is also a heating 20 A slight attenuation change in the bulk Si2p, which suggests a small increase in SiOx formation. The change in peak intensities ˚. suggests an equivalent SiOx thickness increase around 2.5 A Based on the qualitative comparison illustrated in Fig. 1, it is unclear if the SiOx formation is quantitatively reliable. However, the combined decrease in MgO and increase in SiOx does allude to a possible etching mechanism with interface breakdown during oxygen exposure at elevated temperature. To further test for a possible etching mechanism, samples 3 and 4 were heated to 650 8C and exposed to oxygen plasma of 35 mVand 125 mV, respectively, at 5.0 106 Torr for 60 min. After plasma exposure, both samples indicated an overall loss in MgO thickness as well as a significant increase in SiOx formation. Sample 3, which was exposed to a 35 mV plasma, ˚ and a SiOx indicated a MgO thickness loss around 6.4 A ˚ formation around 17 A. Sample 4, which was exposed to a 125 mV plasma with a higher atomic oxygen content than the plasma exposed to sample 3, indicated a MgO thickness loss ˚ and a SiOx formation of 22 A ˚ . A comparison of around 8.1 A the calculated MgO and SiOx thicknesses before and after treatment is illustrated in Fig. 2. The error bars displayed represent the inherent 10% error associated with the thickness calculation. The thicknesses are representative of MgO and SiOx thicknesses that were formed during MgO deposition and after exposure treatment. Therefore, the silicate adlayer that is formed during ex situ hydrogen cleaning is not represented.
Fig. 2. Calculated thicknesses for the breakdown of the MgO/SiC interface and formation of the SiOx layer for each of the heat/oxygen exposures.
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
Although heating in vacuum for 90 min did not change either the film thickness or composition, as the intensity of oxygen exposure conditions increased (with ‘‘intensity’’ defined by the amount of reactive atomic oxygen) the overall thickness of the MgO film decreased and the formation of SiOx increased. This suggests a possible etching/formation mechanism where the presence of oxygen, especially atomic oxygen, at elevated temperatures causes the formation of SiOx at the interface and etching of MgO. AFM surface roughness measurements on the plateaus of the stepped films [4] revealed measurable surface roughening with increased active oxygen exposure, suggesting increased etching. The as-deposited MgO films had an average root mean square (rms) roughness over five sampling areas of 0.27 0.02 nm. The sample heated in vacuum (sample 1-post) had an average rms of 0.38 0.03 nm and the sample annealed under the 125 mV plasma (sample 4post) had an rms of 0.66 0.05 nm. For all surface roughness measurements, the rms errors are statistical errors from multiple areas on repeated samples and do not represent any error in absolute resolution. It is important to note that the etching of the film by reactive oxygen is also temperature dependent. Four samples were used to evaluate the effect of substrate temperature on the breakdown ˚ MgO(1 1 1) was deposited on of the interface. Initially, 20 A the four 6H-SiC(0 0 0 1) substrates at 140 8C. Each sample was then exposed to the harsh oxygen plasma, with a photomultiplier reading of 125 mV, for 60 min at different temperatures. The temperatures of interest included 140 8C, 400 8C, 525 8C, and 650 8C. The degree of interface breakdown increased with increasing temperatures, as shown in Fig. 3 by the increase in SiOx formation with increased substrate temperature. Fig. 3 incorporates the SiOx formation as an accumulation of both SiO and SiO2 at binding energies of 102.1 eVand 103.3 eV, respectively. However, it is important to note that with the exception of the sample exposed to oxygen at 650 8C, the SiOx layers consists of only SiO. The SiOx ˚ composition for the sample at 650 8C was equivalent to 14.0 A ˚ SiO and 8.1 A SiO2. Interestingly, no SiO2 formation was observed for the other three samples.
Fig. 3. Interface breakdown between the MgO/SiC increases significantly with increased substrate temperature when exposed to a harsh oxygen plasma for 60 min. The increase in interface breakdown can be associated with an increase in atomic oxygen diffusing along MgO grain boundaries to initiate the breakdown.
3195
The Huttig temperature where surface mobility becomes significant is defined as 1/3 of the melting point. The Huttig temperatures for MgO and SiC are 943.3 8C and 910 8C, respectively. These temperatures are significantly higher than the temperatures used in this study. Thermodynamic calculations using changes in the Gibbs free energy for the formation of SiOx from SiC and MgO constituents [18] suggest a thermodynamically stable MgO/SiC interface, where silicon bound to carbon will not preferentially bond with oxygen already bound to magnesium for all temperatures used in this study. This indicates that the formation of the SiOx layer is not a result of Si scavenging O atoms from the MgO film. Interface stability was supported experimentally when sample 1 was heated to 740 8C for 90 min and did not show any increase in SiOx nor loss of MgO. If silicon could scavenge oxygen from MgO, sample 1-post would have formed a SiOx layer of a thickness comparable to the other heated samples. This is consistent with other stable interfaces between silicon-based semiconductors and magnesium oxides [5,8,19] as well as the importance of oxygen and high temperature for the oxidation of SiC [20–22]. Our work shows that the formation of a SiOx interface requires the presence of active oxygen in order for breakdown of the MgO/SiC interface. In addition, the thermodynamic calculations show that unbound O provides a thermodynamic driving force to breakdown the otherwise stable MgO/SiC system [18]. This suggests that atomic oxygen must first become available at the MgO/SiC interface in order to initiate SiOx formation. Since the SiOx interface forms readily during exposure to oxygen plasma at elevated temperatures, it is possible that active oxygen reaches the interface along grain boundaries in the MgO to initiate the formation of SiOx. One hypothesized scenario is listed in Eqs. (1)–(4), once the unbound O reaches the MgO/SiC interface. SiCðsÞ þ 2OðgÞ ! SiOðsÞ þ COðgÞ
(1)
SiOðsÞ þ MgOðsÞ ! SiO2ðsÞ þ MgðsÞ
(2)
MgðsÞ ! MgðgÞ
(3)
SiOðsÞ þ OðgÞ ! SiO2ðgÞ
(4)
This hypothesized mechanism explains the stability of the interface when annealed under vacuum, as active oxygen is not available at the interface. It explains the minimal interface breakdown during molecular oxygen exposure where the O2 is not active, compared to the light and harsh plasma containing unbound O species. The hypothesis accounts for the appearance of SiO and SiO2 in the Si2p XPS spectra as well as the loss of MgO through the desorption of magnesium metal as a vapor at the high temperatures (step 3). Although the formation of SiO2 is thermodynamically favored over SiO [18], it is likely that under many processing conditions, the availability of atomic oxygen at the interface is insufficient to account for the total formation of SiO2. This would require the SiO to scavenge oxygen from the MgO film, per step 2. This secondary oxygen scavenging step can account for the experimentally observed
3196
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
increase in interface breakdown between the two plasma exposures, where the harsh plasma (i.e. more available atomic oxygen) showed greater interface breakdown and increased loss of MgO film thickness. From Figs. 2 and 3, the increase in interface breakdown is influenced by both the availability of atomic oxygen and substrate temperature. Fig. 3 indicates a kinetically controlled mechanism where the availability of atomic oxygen at the interface is in excess, and the rate of interface breakdown is kinetically controlled by substrate temperature. Fig. 2, which compares the formation of SiOx at 650 8C to increasing atomic oxygen in the vacuum environment, indicates that the rate of interface breakdown is influenced by the availability of atomic oxygen at the interface. Although the observed contributions of both temperature and atomic oxygen support a kinetically controlled mechanism, it is not possible to rule out the influence of the diffusion rate for atomic oxygen along MgO grain boundaries. The MgO film heated in vacuum (sample 1-post) remained chemically stable based on XPS characterization of the Mg2p photoelectron peak before and after each exposure. In addition, RHEED indicated a slight 3D to 2D transition after the substrate temperature reached 650 8C. When the heated sample was cooled to room temperature, the more 2D RHEED pattern was maintained. A typical RHEED pattern for a MgO film deposited at 140 8C is represented in Fig. 4b. The RHEED pattern is characteristic of MgO(1 1 1) taken along the < 1 1¯ 0 > azimuth and shows some 3D characteristics, as indicated by the intensity spots located in the zeroth Laue zone. During the temperature exposure step of sample 1 to 740 8C in vacuum, the 3D RHEED characteristics were observed to transition to 2D characteristics when the substrate temperature reached 650 8C. The film maintained the (1 1 1) orientation throughout the high-temperature exposure. The transition to 2D characteristics is indicated by the smoother streaks in the zeroth Laue zone, represented in Fig. 4c. This apparent RHEED transition at 650 8C was observed for all four samples that were exposed to different temperature and oxygen conditions. To ˚ MgO(1 1 1) was further explore this transition, a 20 A deposited on 6H-SiC(0 0 0 1) at 140 8C followed by rapidly heating the sample to 650 8C at 1.0 109 Torr. AFM ˚ MgO(1 1 1) film characterization was performed on the 20 A ˚ as-deposited after heating, and was compared to another 20 A film. Based on the 3D characteristics observed by RHEED, it
was expected that the as-deposited film would show higher surface roughness compared to the heated sample. The average surface roughness from five locations on the as-deposited MgO(1 1 1) and the heated MgO(1 1 1) was measured to be 0.27 0.02 nm and 0.26 0.03 nm, respectively. Thus no measurable difference in surface roughness was observed within the resolution limitations of the AFM. No mention has been made up to this point regarding the termination of the MgO(1 1 1) surface. Due to the nature of the (1 1 1) orientation of ionic rocksalt structures, including MgO, the alternating layers of oppositely charged magnesium and oxygen result in a dipole moment perpendicular to the surface. In order to minimize this instability, the surface cannot truly be terminated by a layer of oxygen or magnesium. Much research has been conducted to understand the true surface termination of MgO(1 1 1) [23–25,29]. Early work suggested the surface became faceted into neutral {1 0 0} planes [23,24]. However, more recent research has reported that faceting is not a favorable means of stabilization, but rather MgO(1 1 1) surface terminations of H3 H3 R308, 2H3 2H3 R308, or 2 2 reconstructions are predicted to be more favorable for minimizing electric dipoles [25–29]. Although these surface reconstructions are electrically stable, they often require high temperatures (>800 8C) to form. Further research combining experimental and theoretical studies supported a hypothesis for an OH terminated MgO(1 1 1)-(1 1) surface, which is most energetically favorable and results in a surface termination that is more stable than the non-polar MgO(1 0 0) surface [29–31]. Through careful characterization of our MgO films by XPS ˚ and RHEED we see evidence of this OH termination. A 100 A MgO(1 1 1) film was deposited on hydrogen cleaned 6HSiC(0 0 0 1) at a substrate temperature of 140 8C. A thickness of ˚ was chosen to ensure that the oxygen O1s photoelectron 100 A signal was originating from only the MgO film and not the ˚ is greater than the silicate adlayer on the 6H-SiC, since 100 A attenuation depth of electrons originating from oxygen at the interface. The resulting RHEED pattern indicated some 3D characteristics, as expected, but maintained a pattern characteristic of MgO(1 1 1). XPS characterization of the film indicated one Mg2p photoelectron peak, but two O1s photoelectron peaks. The primary O1s peak, associated with MgO, was located at a binding energy around 530 eV. The secondary peak was located at a binding energy 2 eV higher and was designated as the OH surface state, which is in close agreement with Lazarov et al. [31].
Fig. 4. RHEED patterns for MgO growth and heat exposure illustrating the heat-induced transition from 3D characteristics to 2D characteristics. (a) Hydrogen ˚ MgO(1 1 1) deposited on hydrogen cleaned 6H-SiC(0 0 0 1) at 140 8C, and (c) 20 A ˚ MgO(1 1 1) film cleaned 6H-SiC(0 0 0 1), (b) heteroepitaxial growth of 20 A annealed rapidly to 650 8C at 1.0 109 Torr and held for 2 min.
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
Taking into account sampling volume, the relative intensity of the OH peak to the MgO peak was calculated to be 0.11, indicating slightly more than one monolayer coverage. Quantitative characterization of the actual monolayer coverage would require the use of angle resolved XPS. However, this was not available, leaving the relative ratio of OH to MgO as a qualitative measure. The monolayer coverage of OH greater than 1 is most likely due to the 3D features observed by RHEED, which would increase the overall surface area and therefore increase the amount of available sites for OH termination. ˚ MgO film was grown at 140 8C and was then A 100 A annealed to 650 8C at 1.0 109 Torr for 2 min in order to transition to the more 2D surface, as observed through RHEED. The more 2D RHEED pattern was observed on the 650 8C film as well as when the film had cooled to room temperature. XPS characterization of the film indicated a significant drop in the relative amount of OH to MgO. On the annealed film, the ratio of OH to MgO, as indicated by the O1s peak, was calculated to be 0.08. The decrease in the relative amount of OH is most likely due to a surface smoothing effect. Additionally, a third sample was deposited to a thickness of ˚ at a 650 8C growth temperature in order to maintain a 100 A smooth 2D surface during growth and then to evaluate the surface termination after cooling. XPS characterization of the film indicated a relative OH to MgO ratio of 0.06 as calculated from the O1s peak. The additional decrease in the relative ratio can again be associated with the smoother 2D surface that was maintained during film growth. Fig. 5 illustrates the O1s ˚ low-temperature MgO, photoelectron spectra for the 100 A ˚ annealed MgO, and 100 A high-temperature MgO. XPS characterization of all the films discussed indicate a MgO(1 1 1)-(1 1) OH terminated film. The decrease in OH to MgO is most likely associated with a smoother MgO surface, which was best observed by RHEED.
˚ MgO(1 1 1). Bottom spectra is for MgO Fig. 5. XPS spectra of O1s for 100 A as-deposited at 140 8C, middle spectra is the same MgO film annealed to 650 8C for 2 min, and top spectra is Mg-O deposited at 650 8C. The decrease in Mg-OH is believed to be related to the observed RHEED transition from 3D to 2D, which results in a smoother surface.
3197
The smoother MgO surface resulted in less 3D features and therefore less surface area to be terminated by the OH group. It is currently unclear if the decrease in 3D features is related to a decrease in grain boundaries or only surface smoothing. If the MgO film deposited at 650 8C does in fact have fewer grain boundaries, then this may help better understand the contribution of oxygen diffusion along grain boundaries, which is being considered for future in-depth understanding of the interface breakdown mechanism. Although the MgO film deposited on the 6H-SiC at 650 8C was too thick to observe any interface breakdown by XPS, similar high temperature growths have ˚ . Interestbeen performed with MgO thicknesses around 20 A ingly, although the substrate was shuttered from the line-of-site of the oxygen source, even at substrate temperature of 650 8C and an oxygen environment of 5.0 106 Torr and 35 mV there was no evidence by XPS or RHEED of surface oxidation prior to growth. This may be due to the thermodynamically favored formation of MgO compared to SiO. Although SiO2 is thermodynamically favorable over both SiO and MgO, it does require bonds to be broken within the silicate adlayer of the SiC. Therefore, it is not too surprising that the formation of MgO is favorable given the availability of atomic magnesium and atomic oxygen at the surface. Degradation of the MgO/SiC interface may be detrimental to ultimate device performance. In order to maintain an effective MgO/SiC interface, the subsequent deposition of the functional oxide should avoid long-term exposure of the MgO layer to temperatures above 650 8C in an oxygen environment, especially an active oxygen environment such as the oxygen plasmas often used in MBE depositions. To test the feasibility and effectiveness of MgO(1 1 1) as a template layer for functional oxides, BTO was deposited by MBE on MgO/SiC and bare SiC. When deposited on the bare 6H-SiC, the resulting BTO was amorphous. Although no interface breakdown or oxidation of silicon was observed, the H3 H3 R308 surface of the 6H-SiC(0 0 0 1) did not promote the heteroepitaxy of BTO. Based on the crystal structure and lattice spacing of the 6H-SiC surface reconstruction and the O–O spacing of the BTO(1 1 1) there are two likely alignments. One has a compressive mismatch of 6.2% with a 308 in-plane rotation of the BTO(1 1 1) on the 6H-SiC silicate adlayer. The second has a tensile mismatch of 7.9% with no inplane rotation, but does require two atomic spacings of BTO(1 1 1) to one reconstructed 6H-SiC. Fig. 6a and b show the RHEED patters for the BTO films deposited on bare 6H-SiC. ˚ MgO template layer under When BTO was deposited on the 20 A the same growth conditions, the resulting BTO was crystalline with a distinct, transmission dominated RHEED pattern characteristic of a twinned (1 1 1) oriented tetragonal perovskite structure (Fig. 6e). XRD of the BTO film resulted in only (1 1 1) orientated BTO. The observed heteroepitaxy of the BTO on the MgO template layer is most likely do to the improved lattice matching. The calculated lattice mismatch based on the observed epitaxial relationship and the O–O spacing for both MgO(1 1 1) and BTO(1 1 1) is 4.4% in tension. The decrease in tensile mismatch from 7.9%, for BTO/6H-SiC, to 4.4%, for BTO/MgO/ 6H-SiC promoted the heteroepitaxial integration of BTO on 6HSiC. This demonstrates the importance and effectiveness of
3198
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
Fig. 6. RHEED patterns illustrating the importance of the MgO(1 1 1) template layer for integrating crystalline BTO on 6H-SiC(0 0 0 1). The crystalline MgO(1 1 1) layers proved essential for the heteroepitaxy of the BTO film.
crystalline MgO(1 1 1) as a heteroepitaxial template layer for integrating a wide variety of (1 1 1) oriented perovskite ferroelectrics on the wide band gap semiconductor 6H-SiC.
thank Dr. Al Sacco at Northeastern University for use of the AFM, and Lena Fitting Kourkoutis at Cornell University for independent confirmation of film thickness by TEM.
4. Summary and conclusions References The post-deposition effects of heating and oxygen species on chemistry, morphology, and crystallography of MgO(1 1 1) on ˚ ) MgO films were 6H-SiC(0 0 0 1) were evaluated. Thin (20 A found to be thermally stable when heated to 740 8C at 1.0 109 Torr, and when exposed to a harsh plasma pressure of 5.0 106 Torr at 140 8C. However, interfacial breakdown and surface roughening began to occur when the films were exposed to molecular oxygen or oxygen plasma at 650 8C for 60 min. Exposure to a constant harsh plasma at different temperatures indicated a kinetic component to the interface breakdown where the rate of interface breakdown was found to increase significantly at temperatures above 400 8C. Also occurring at 650 8C was a transition from a 3D to more 2D RHEED pattern and a reduction in OH surface termination in the MgO film, most likely due to the decrease in surface area of the smoother MgO. Despite these changes, RHEED patterns indicated the preservation of the MgO(1 1 1) orientation under all conditions. Successful deposition of crystalline BTO(1 1 1) was obtained by MBE when deposited on the MgO/6H-SiC, but not on the hydrogen cleaned 6H-SiC. The improved epitaxy demonstrates the potential for using crystalline MgO as a heteroepitaxial template layer for integrating a wide range of perovskite ferroelectrics on 6H-SiC. Acknowledgements This research was supported by Office of Naval Research under contracts N0014-06-1-0761 and the N00014-04-1-0426, Dr. Colin Wood Program Monitor. The authors would like to
[1] W.A. Doolittle, A.G. Carver, W. Henderson, J. Vac. Sci. Technol. B 23 (2005) 1272. [2] Z. Chen, A. Yang, S.D. Yoon, K. Ziemer, C. Vittoria, V.G. Harris, J. Magn. Magn. Mater. 301 (2006) 166. [3] A.R. Meier, F. Niu, B.W. Wessels, J. Cryst. Growth 294 (2006) 401. [4] T.L. Goodrich, J. Parisi, Z. Cai, K.S. Ziemer, Appl. Phys. Lett. 90 (2007) 042910. [5] T.L. Goodrich, Z. Cai, M.D. Losego, J.-P. Maria, K.S. Ziemer, J. Vac. Sci. Technol. B 25 (2007) 1033. [6] J.-J. Chen, B.P. Gila, M. Hlad, A. Gerger, F. Ren, C.R. Abernathy, S.J. Pearton, Appl. Phys Lett. 88 (2006) 042113. [7] H.S. Craft, J.F. Ihlefeld, M.D. Losego, R. Collazo, Z. Sitar, J.-P. Maria, Appl. Phys. Lett. 88 (2006) 212906. [8] F. Niu, B.H. Hoerman, B.W. Wessels, J. Vac. Sci. Technol. B 18 (2000) 2146. [9] J. Zhu, X.H. Wei, Y. Zhang, Y.R. Li, J. Appl. Phys. 100 (2006) 104106. [10] J.H. Haeni, et al. Lett. Nat. 430 (2004) 758. [11] K.A. Shaw, E. Lochner, D.M. Lind, J. Appl. Phys. 87 (2000) 1727. [12] J. Bernhardt, J. Schardt, U. Starke, K. Heinz, App. Phys. Lett. 74 (1999) 1084. [13] T. Goodrich, J. Parisi, J. Leong, K.S. S Ziemer, AIChE Annual Meeting Conference Proceedings, Cincinnati, OH, United States, Oct. 30–Nov. 4, (2005), p. 135e/1. [14] C.D. Theis, D.G. Schlom, J. Vac. Sci. Technol. A 14 (1996) 2677. [15] NIST Database 82, U.S. Department of Commerce (2001). [16] Y.R. Li, Z. Liang, Y. Zhang, J. Zhu, S.W. Jiang, X.H. Wei, Thin Solid Films 489 (2005) 245. [17] http://srdata.nist.gov/xps. [18] M.W. Chase, Jr., C.A. Davies, J.R. Downey, Jr., D.J. Frurip, R.A. McDonald, A.N. Syverud, J. Phys. Chem. Ref. Data 14 (1985) JANAF Thermochemical Tables. [19] X.Y. Chen, K.H. Wong, C.L. Mak, J.M. Liu, X.B. Yin, M. Wang, Z.G. Liu, Appl. Phys. A 74 (2002) 703. [20] D. Schmeiber, D.R. Batchelor, R.P. Mikalo, P. Hoffmann, A. Lloyd-Spetz, Appl. Surf. Sci. 184 (2001) 340.
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199 [21] G.G. Jernigan, R.E. Stahlbush, N.S. Saks, Appl. Phys. Lett. 77 (2000) 1437. [22] W. Chen, X.N. Xie, H. Xu, A.T.S. Wee, K.P. Loh, J. Phys. Chem. B 107 (2003) 11597. [23] V.E. Henrich, Surf. Sci. 57 (1976) 385. [24] H. Onishi, C. Egawa, T. Aruga, Y. Iwasawa, Surf. Sci. 191 (1987) 479. [25] R. Plass, K. Egan, C. Collazo-Davila, D. Grozea, E. Landree, L.D. Marks, M. Gajdardziska-Josifovska, Phys. Rev. Lett. 81 (1998) 4891. [26] D. Wolf, Phys. Rev. Lett. 68 (1992) 3315.
3199
[27] F. Finocchi, A. Barbier, J. Jupille, C. Noguera, Phys. Rev. Lett. 92 (2004) 136101. [28] A. Subramanian, L.D. Marks, O. Warschkow, D.E. Ellis, Phys. Rev. Lett. 92 (2004) 026101. [29] A. Wander, I.J. Bush, N.M. Harrison, Phys. Rev. B 68 (2003) 233405. [30] K. Refson, R.A. Wogelius, D.G. Fraser, M.C. Payne, M.H. Lee, V. Milman, Phys. Rev. B 52 (1995) 10823. [31] V.K. Lazarov, R. Plass, H.-C. Poon, D.K. Saldin, M. Weinert, S.A. Chamber, M. Gajdardziska-Josifovska, Phys. Rev. B 71 (2005) 115434.
Applied Surface Science 254 (2008) 3191–3199 www.elsevier.com/locate/apsusc
Stability of MgO(1 1 1) films grown on 6H-SiC(0 0 0 1) by molecular beam epitaxy for two-step integration of functional oxides T.L. Goodrich, Z. Cai, K.S. Ziemer * Department of Chemical Engineering, Northeastern University, Boston, MA 02115, United States Received 25 July 2007; received in revised form 27 October 2007; accepted 30 October 2007 Available online 4 November 2007
Abstract ˚ thick, was grown by molecular beam epitaxy (MBE) on hydrogen cleaned hexagonal silicon Crystalline magnesium oxide (MgO) (1 1 1), 20 A carbide (6H-SiC). The films were further heated to 740 8C and 650 8C under different oxygen environments in order to simulate processing conditions for subsequent functional oxide growth. The purpose of this study was to determine the effectiveness and stability of crystalline MgO films and the MgO/6H-SiC interface for subsequent heteroepitaxial deposition of multi-component, functional oxides by MBE or pulsed laser deposition processes. The stability of the MgO films and the MgO/6H-SiC interface was found to be dependent on substrate temperature and the presence of atomic oxygen. The MgO films and the MgO/6H-SiC interface are stable at temperatures up to 740 8C at 1.0 109 Torr for extended periods of time. While at temperatures below 400 8C exposure to the presence of active oxygen for extended periods of time has negligible impact, exposure to the presence of active oxygen for more than 5 min at 650 8C will degrade the MgO/6H-SiC interface. Concurrent etching and interface breakdown mechanisms are hypothesized to explain the observed effects. Further, barium titanate was deposited by MBE on bare 6H-SiC(0 0 0 1) and MgO(1 1 1)/6H-SiC(0 0 0 1) in order to evaluate the effectiveness of the MgO as a heteroepitaxial template layer for perovskite ferroelectrics. # 2007 Elsevier B.V. All rights reserved. PACS : 81.15.Hi 79.60.i Keywords: MgO; SiC; Heteroepitaxy; Functional oxides; MBE; Wide bandgap semiconductor
1. Introduction The realization of next-generation devices for hightemperature, high-frequency applications will require the integration of multi-component, functional oxides on wide bandgap semiconductors, including hexagonal silicon carbide (6H-SiC) [1]. Multi-component, functional oxides, including multiferroics (bismuth ferrite), ferroelectrics (barium titanate), piezoelectrics (lead zirconate titanate), and ferrimagnetics (barium hexaferrite) are of interest because of the potential for controlled coupling between mechanical, electrical, and magnetic properties. In order for the successful integration of functional oxides on 6H-SiC, it is necessary to establish an abrupt and effective interface between the 6H-SiC and the functional oxide film. Chen et al. [2] reported on the improved crystallography and magnetic properties of M-type barium
* Corresponding author. Tel.: +1 617 373 2990; fax: +1 617 373 2209. E-mail address: [email protected] (K.S. Ziemer). 0169-4332/$ – see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2007.10.077
hexaferrite (BaM, BaFe12O19) grown by pulsed laser deposition (PLD) on 6H-SiC(0 0 0 1) through the use of a graduated magnesium oxide (MgO) interlayer. Meier et al. [3] demonstrated improved epitaxial growth of (0 0 1) oriented barium titanate (BTO) grown by metal organic chemical vapor deposition by integrating crystalline MgO(0 0 1) through molecular beam epitaxy (MBE) on epitaxial strontium titanate ˚ crystalline on Si(0 0 1). We have previously shown that a 20 A MgO(1 1 1) film deposited on 6H-SiC(0 0 0 1) by MBE provides an effective heteroepitaxial template and prevents interface breakdown during functional oxide film growth [4,5]. The bulk lattice spacing of MgO(1 1 1) suggests that the film acts as a buffer layer to improve lattice matching between the 6HSiC substrate and the desired functional oxide. For example, 6H˚ compared to the SiC(0 0 0 1) has a bulk lattice spacing of 3.08 A oxygen to oxygen lattice spacing of BaM(0 0 0 1), which is ˚ . The oxygen to oxygen lattice spacing of MgO(1 1 1) is 2.94 A ˚ , which when integrated as a heteroepitaxial template, can 2.98 A reduce lattice mismatch and aid in creating an effective interface through controlled heteroepitaxy. In order for the MgO layer to
3192
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
be an effective heteroepitaxial template, it is necessary to deposit smooth, single crystalline MgO(1 1 1) on the 6H-SiC(0 0 0 1) substrate. Several groups have demonstrated the ability to grow smooth crystalline MgO(1 1 1) or MgO(1 0 0) on the semiconductor substrates of gallium nitride [6,7], silicon [3], and silicon carbide [4,5,8] by MBE. Typical deposition conditions include temperatures ranging from room temperature up to 600 8C. For use of the MgO film as an interlayer between a ferroelectric, ferromagnetic, or multiferroic oxide, the substrate and MgO film may be exposed to different environmental conditions from the MgO film deposition conditions. In the MBE environment, this typically means temperatures up to 700 8C and active oxygen species from a plasma source or an ozone source [9–11]. Usually, however, the functional oxides are deposited through a physical deposition process, like PLD where the MgO/ SiC would be exposed to growth temperatures as high as 925 8C and even higher annealing temperatures, which can exceed 1000 8C, and millitorr or atmospheric pressures of oxygen [2]. Thus a variety of processing conditions, and perhaps two processing technologies (MBE and PLD), may be necessary to realize full integration of device quality functional oxides and semiconductors. This study takes the next step in realizing a twostep growth process by understanding the impact of different oxide processing environments on thin MgO films and the MgO(1 1 1)/SiC(0 0 0 1) interface. In this work, MgO thin films were deposited on 6HSiC(0 0 0 1) at a constant substrate temperature of 140 8C and an oxygen plasma at 5.0 106 Torr with a photomultiplier reading of 35 mV. These processing conditions result in an adsorption controlled growth regime where the growth rate of the MgO film increases linearly with respect to magnesium flux ˚ MgO films were previously found to be thermally [5]. The 20 A stable as indicated by X-ray photoelectron spectroscopy (XPS), reflection high-energy electron diffraction (RHEED), and atomic force microscopy (AFM), after heating to 740 8C at ˚ 1.0 109 Torr for 90 min. Similarly, heating 380 A MgO(1 1 1) films deposited on 6H-SiC(0 0 0 1) to 790 8C for 30 min in vacuum and in air were again thermally stable, as indicated by XPS, RHEED, AFM, and a Bruker D-5000 diffractometer with Highstar area detection [5]. Due to the ˚ MgO films, it was not possible to thickness of the 380 A characterize the interface by XPS, as the XPS sampling depth ˚ . Therefore, X-ray for Si and C through MgO is 93 A diffraction (XRD) was performed on the films. The XRD patterns contained diffraction peaks characteristic of only 6HSiC(0 0 0 1) and MgO(1 1 1), and did not indicate any additional phases or breakdown in the MgO/SiC after exposure. This paper reports the effects of post-deposition temperature and oxygen species exposure on the MgO(1 1 1) film and the MgO(1 1 1)/SiC(0 0 0 1) interface. Chemical, morphological, and crystallographic properties are compared before and after the exposures, which were chosen based on the typical processing requirements for subsequent MBE or PLD growth of multi-component functional oxides, but under extended time periods. The purpose of this extended time exposure was to understand the impact of a subsequent deposition environment on the MgO layers. As shown in this paper, certain
combinations of elevated temperature and oxygen species can detrimentally impact the film and interface. By understanding potential mechanisms of degradation, successful twostep growth processes can be developed in order to avoid conditions that result in interface breakdown during actual deposition of the multilayered heterostructure. 2. Experimental 6H-SiC(0 0 0 1) (Cree, Inc.) substrates were degreased in heated solvents of trichloroethylene, acetone, and methanol. The samples were then cleaned in a custom built hydrogen flow furnace in order to remove surface contamination and scratches resulting from mechanical polishing. The cleaned 6HSiC(0 0 0 1) surface consisted of a H3 H3 R30o silicate adlayer surface reconstruction, similar to the results reported by Bernhardt et al. [12]. Details and characterization of the clean 6H-SiC(0 0 0 1) surface are described elsewhere [4,5,13]. The samples were then mounted on a molybdenum puck with conductive silver paint, immediately loaded into a ultra-high vacuum system, and evacuated to a base pressure of 1.0 109 Torr. Chemical and bonding characterization was performed by XPS using a PHI model 04-173-0-007 Mg/Al dual anode, nonmonochromatic X-ray source and a PHI model 10-360-4-015 hemispherical analyzer. XPS peak deconvolution was performed using a minimum full width at half maximum of 1.6 eV and an 80/20 Gaussian/Lorenzian peak shape, as determined from a clean Au4f7 photoelectron peak. In situ surface crystallographic characterization was performed using a Staib Instruments RHEED system RH-15 and a k-Space Associates KSA 400 digital camera and software. Ex situ morphological characterization was performed using a NANOPROBE IIIA Digital Instruments AFM in tapping mode with an oxide sharpened SiN tip with a radius of curvature less than 10 nm. MgO was deposited by MBE using a SPECS Scientific Instruments dual source, low-temperature effusion cell filled with magnesium shavings (99.98%, Alpha Aesar) and an Oxford Applied Research remote oxygen rf-plasma source, model HD25. The plasma source was equipped with ion deflection plates located at the end of the discharge tube, which prevented any charged species from impinging on the surface during deposition. The absence of ions was confirmed by setting the ionization energy to zero on a Hiden Analytical HALO 201 quadrupole mass spectrometer, and turning the ion deflection plates of the plasma source on and off. The plasma source was also equipped with an optical photodiode filtered for a wavelength of 844 nm, which allowed direct evaluation of the atomic oxygen content in the plasma. Substrate temperature was measured using a two-color optical pyrometer and a type C thermocouple in contact with the backside of the molybdenum sample holder. In all experiments, the MgO films were deposited in a constant oxygen plasma environment with a background pressure of 5.0 106 Torr and a filtered photomultiplier reading of 35 mV. Although quantitative characterization of the oxygen plasma has not been performed, the optical photodiode
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
filter on the plasma discharge tube enables a qualitative measure of atomic oxygen species. An increase in atomic oxygen is directly associated with an increase in the photomultiplier reading. Through preliminary characterization, the maximum obtainable photomultiplier reading at this oxygen throughput was found to be around 400 mV. Comparing the photomultiplier reading of 35 mV used in this study with the maximum obtainable reading of 400 mV indicates that the atomic oxygen content of the oxygen plasma was relatively low. In all cases, the plasma source was operated with a voltage bias between the two deflection plates at the exit of the discharge tube, thus removing all oxygen ions in the plasma before they could reach the substrate surface. BTO films were deposited by MBE at substrate temperatures of 650 8C and constant oxygen plasma at 5.0 106 Torr with a filtered photomultiplier reading of 35 mV. The barium was supplied from a barium rod (99.9%, Electronic Space Products International) in the same SPECS Scientific Instruments dual source, low-temperature effusion cell as the magnesium. The dual source effusion cell consists of two independently controlled cells capable of supplying two different materials. In our system the source is set up with magnesium and barium. The titanium was supplied by a Varian Inc. mini Ti-Ball source, originally designed as a sublimation source for an ion pump, and is based off the design by Theis et al. [14]. Film thicknesses were calculated using the TPP-2M and Gries equations [15] and the XPS signal attenuation of the Si2p photoelectrons. These thickness calculations have an inherent error around 10% due to transport approximations, surface roughness, surface excitations, and surface refraction [15]. For the material parameters used in these equations, all thickness calculations were based on the physical properties of bulk, crystalline materials. It is fairly safe to assume bulk properties for all MgO film thickness calculations used in this study based on the findings by Li et al. [16], who reported strain relaxation of the crystal lattice to bulk properties for MgO(1 0 0) deposited on SrTiO3(1 0 0) at thickness of 1.75 monolayers ˚ . Secondary thickness measurements were carried out or 7.4 A ˚ MgO film, using transmission electron microscopy on a 25 A which confirmed the reliability of the attenuation thickness calculation within a 10% error. For some cases described in the results and discussion section below, a layered structure of MgO/SiOx/SiC is formed. In order to
3193
quantitatively analyze the formation of the SiOx layer, modified thickness calculations were necessary to take into consideration appropriate attenuation parameters due to the hypothesized MgO/SiOx/SiC layered structure. As described later, the assumption that the SiOx is located at the interface with negligible intermixing between the MgO and SiC was supported by XPS and RHEED, which indicated the crystalline MgO(1 1 1) pattern was maintained on the surface during all exposures. Since the MgO is assumed to be on the surface, the thickness of the MgO film after exposure was determined through a direct proportion between the area of the deconvoluted XPS Mg2p peaks and the calculated thicknesses of the as-deposited film. The overall attenuation of the silicon was known by comparing the XPS intensity before growth, after growth, and after exposure. Similarly, the attenuation of silicon through SiOx and MgO was determined as a function of varying thickness. Since the thickness of MgO after exposure was proportionally calculated, the thickness of the SiOx layer was iterated until the overall observed silicon attenuation was achieved. Note that the assumption of a layered structure increases the error associated with the thickness calculation. Although accurate thickness quantification is very difficult by this method, the associated error is consistent for each sample, which allows for qualitative comparison between the different exposure conditions. Thus arguments are made based on a relative comparison in order to support the qualitative analysis of undesirable interface degradation between the MgO and 6H-SiC. 3. Results and discussion ˚ of MgO(1 1 1) was In the first set of experiments, 20 A deposited on hydrogen cleaned 6H-SiC(0 0 0 1) at a substrate temperature of 140 8C, a magnesium flux of 1.0 1014 cm2 s1, an oxygen chamber pressure of 5.0 106 Torr, and a photomultiplier reading of 35 mV. These MgO films were then exposed to four different post-deposition conditions: (1) heated to 740 8C in vacuum (1.0 109 Torr) for 90 min, (2) heated to 650 8C in molecular oxygen (5.0 106 Torr) for 60 min, (3) heated to 650 8C in a light oxygen plasma (5.0 106 Torr, 35 mV) for 60 min, and (4) heated to 650 8C in a harsh plasma (5.0 106 Torr, 125 mV) for 60 min. The processing conditions for the as-deposited films and post-deposition exposure are listed in Table 1. RHEED was
Table 1 Sample identification with the processing parameters used Sample
Time (min)
Temperature (8C)
Mg flux (cm2 s1)
O2 pressure (Torr)
Photomultiplier (mV)
14
6
5.0 10 1.0 109
35 0
1
Pre Post
5 90
140 740
1.0 10 N/A
2
Pre Post
5 60
140 650
1.0 1014 N/A
5.0 106 5.0 106
35 0
3
Pre Post
5 60
140 650
1.0 1014 N/A
5.0 106 5.0 106
35 35
4
Pre Post
5 60
140 650
1.0 1014 N/A
5.0 106 5.0 106
35 125
Samples 1–4 represent the four as-deposited MgO films (pre) and post-deposition long-term exposures (post).
3194
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
used real-time to monitor the crystal orientation of the MgO films during growth and during post-deposition exposures. The four as-deposited MgO films (samples 1-pre, 2-pre, 3-pre, ˚. and 4-pre) were grown to an average thickness of 20.3 2.9 A ˚ The 2.9 A does not include the 10% error associated with the thickness calculations. Rather, it is indicative of the error associated with thickness variation between the four samples, ˚ . XPS which were grown to a target thickness of 20 A characterization of the four as-deposited MgO films indicated that they were stoichiometric with an average composition of MgO1.070.18. After each post-deposition exposure, XPS analysis indicated that the chemistry for all of the MgO films (samples 1-post, 2-post, 3-post, and 4-post) were stable with no change in bonding state of the Mg2p peak and no indication of magnesium peroxide, magnesium metal, or magnesium silicide. However, after exposure to oxygen at elevated temperature (samples 2-post, 3-post, and 4-post) the O1s and Si2p photoelectron peaks indicated the formation of a SiOx layer. Fig. 1 qualitatively illustrates the appearance of the SiOx layer as indicated by the Si2p XPS spectra. The peak intensities in Fig. 1 have not been normalized, and thus illustrate the additional signal attenuation of the silicon in bulk SiC through the SiOx layer. This additional attenuation was used to help calculate the thickness of the SiOx layer and follows the assumption that the SiOx layer is located at the MgO/SiC interface. RHEED indicated the crystalline MgO(1 1 1) pattern was maintained on the surface during all exposures. Thus, it is reasonable to assume that the SiOx is forming at the MgO/SiC interface and does not influence the crystal structure of the MgO film. The resulting MgO/SiOx/SiC layered structure is most evident in samples 3-post and 4-post. The SiOx layer is composed of two different oxidation states of the silicon. The left most peak, with a binding energy around 103.3 eV is associated with SiO2, while the middle peak, with a binding energy around 102.1 eV, is associated with SiO [17].
Fig. 1. XPS spectra of Si2p for samples 1–4 after heat/oxygen exposure illustrating the increased MgO/SiC interface breakdown with exposure to increased atomic oxygen. The SiOx formation for samples 3-post and 4-post consists of both SiO2 and SiO.
Sample 1, which was heated to 740 8C for 90 min at 1.0 109 Torr with no oxygen exposure, showed a slight ˚ . This change decrease in MgO film thickness of less than 0.2 A in thickness is within the error of the thickness calculation itself and is not considered significant. In addition, the attenuated signal originating from the silicate adlayer remained unchanged after exposure. This illustrates the thermal stability of the MgO/SiC interface when heated in vacuum. The MgO film of sample 2, which was heated to 650 8C under molecular oxygen at a pressure of 5.0 106 Torr for ˚ . This is a 15% decrease 60 min, decreased in thickness by 2.6 A in MgO thickness compared to the as-deposited film, which is outside the 10% calculation error and is a repeatable trend when ˚ MgO films in molecular oxygen. There is also a heating 20 A slight attenuation change in the bulk Si2p, which suggests a small increase in SiOx formation. The change in peak intensities ˚. suggests an equivalent SiOx thickness increase around 2.5 A Based on the qualitative comparison illustrated in Fig. 1, it is unclear if the SiOx formation is quantitatively reliable. However, the combined decrease in MgO and increase in SiOx does allude to a possible etching mechanism with interface breakdown during oxygen exposure at elevated temperature. To further test for a possible etching mechanism, samples 3 and 4 were heated to 650 8C and exposed to oxygen plasma of 35 mVand 125 mV, respectively, at 5.0 106 Torr for 60 min. After plasma exposure, both samples indicated an overall loss in MgO thickness as well as a significant increase in SiOx formation. Sample 3, which was exposed to a 35 mV plasma, ˚ and a SiOx indicated a MgO thickness loss around 6.4 A ˚ formation around 17 A. Sample 4, which was exposed to a 125 mV plasma with a higher atomic oxygen content than the plasma exposed to sample 3, indicated a MgO thickness loss ˚ and a SiOx formation of 22 A ˚ . A comparison of around 8.1 A the calculated MgO and SiOx thicknesses before and after treatment is illustrated in Fig. 2. The error bars displayed represent the inherent 10% error associated with the thickness calculation. The thicknesses are representative of MgO and SiOx thicknesses that were formed during MgO deposition and after exposure treatment. Therefore, the silicate adlayer that is formed during ex situ hydrogen cleaning is not represented.
Fig. 2. Calculated thicknesses for the breakdown of the MgO/SiC interface and formation of the SiOx layer for each of the heat/oxygen exposures.
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
Although heating in vacuum for 90 min did not change either the film thickness or composition, as the intensity of oxygen exposure conditions increased (with ‘‘intensity’’ defined by the amount of reactive atomic oxygen) the overall thickness of the MgO film decreased and the formation of SiOx increased. This suggests a possible etching/formation mechanism where the presence of oxygen, especially atomic oxygen, at elevated temperatures causes the formation of SiOx at the interface and etching of MgO. AFM surface roughness measurements on the plateaus of the stepped films [4] revealed measurable surface roughening with increased active oxygen exposure, suggesting increased etching. The as-deposited MgO films had an average root mean square (rms) roughness over five sampling areas of 0.27 0.02 nm. The sample heated in vacuum (sample 1-post) had an average rms of 0.38 0.03 nm and the sample annealed under the 125 mV plasma (sample 4post) had an rms of 0.66 0.05 nm. For all surface roughness measurements, the rms errors are statistical errors from multiple areas on repeated samples and do not represent any error in absolute resolution. It is important to note that the etching of the film by reactive oxygen is also temperature dependent. Four samples were used to evaluate the effect of substrate temperature on the breakdown ˚ MgO(1 1 1) was deposited on of the interface. Initially, 20 A the four 6H-SiC(0 0 0 1) substrates at 140 8C. Each sample was then exposed to the harsh oxygen plasma, with a photomultiplier reading of 125 mV, for 60 min at different temperatures. The temperatures of interest included 140 8C, 400 8C, 525 8C, and 650 8C. The degree of interface breakdown increased with increasing temperatures, as shown in Fig. 3 by the increase in SiOx formation with increased substrate temperature. Fig. 3 incorporates the SiOx formation as an accumulation of both SiO and SiO2 at binding energies of 102.1 eVand 103.3 eV, respectively. However, it is important to note that with the exception of the sample exposed to oxygen at 650 8C, the SiOx layers consists of only SiO. The SiOx ˚ composition for the sample at 650 8C was equivalent to 14.0 A ˚ SiO and 8.1 A SiO2. Interestingly, no SiO2 formation was observed for the other three samples.
Fig. 3. Interface breakdown between the MgO/SiC increases significantly with increased substrate temperature when exposed to a harsh oxygen plasma for 60 min. The increase in interface breakdown can be associated with an increase in atomic oxygen diffusing along MgO grain boundaries to initiate the breakdown.
3195
The Huttig temperature where surface mobility becomes significant is defined as 1/3 of the melting point. The Huttig temperatures for MgO and SiC are 943.3 8C and 910 8C, respectively. These temperatures are significantly higher than the temperatures used in this study. Thermodynamic calculations using changes in the Gibbs free energy for the formation of SiOx from SiC and MgO constituents [18] suggest a thermodynamically stable MgO/SiC interface, where silicon bound to carbon will not preferentially bond with oxygen already bound to magnesium for all temperatures used in this study. This indicates that the formation of the SiOx layer is not a result of Si scavenging O atoms from the MgO film. Interface stability was supported experimentally when sample 1 was heated to 740 8C for 90 min and did not show any increase in SiOx nor loss of MgO. If silicon could scavenge oxygen from MgO, sample 1-post would have formed a SiOx layer of a thickness comparable to the other heated samples. This is consistent with other stable interfaces between silicon-based semiconductors and magnesium oxides [5,8,19] as well as the importance of oxygen and high temperature for the oxidation of SiC [20–22]. Our work shows that the formation of a SiOx interface requires the presence of active oxygen in order for breakdown of the MgO/SiC interface. In addition, the thermodynamic calculations show that unbound O provides a thermodynamic driving force to breakdown the otherwise stable MgO/SiC system [18]. This suggests that atomic oxygen must first become available at the MgO/SiC interface in order to initiate SiOx formation. Since the SiOx interface forms readily during exposure to oxygen plasma at elevated temperatures, it is possible that active oxygen reaches the interface along grain boundaries in the MgO to initiate the formation of SiOx. One hypothesized scenario is listed in Eqs. (1)–(4), once the unbound O reaches the MgO/SiC interface. SiCðsÞ þ 2OðgÞ ! SiOðsÞ þ COðgÞ
(1)
SiOðsÞ þ MgOðsÞ ! SiO2ðsÞ þ MgðsÞ
(2)
MgðsÞ ! MgðgÞ
(3)
SiOðsÞ þ OðgÞ ! SiO2ðgÞ
(4)
This hypothesized mechanism explains the stability of the interface when annealed under vacuum, as active oxygen is not available at the interface. It explains the minimal interface breakdown during molecular oxygen exposure where the O2 is not active, compared to the light and harsh plasma containing unbound O species. The hypothesis accounts for the appearance of SiO and SiO2 in the Si2p XPS spectra as well as the loss of MgO through the desorption of magnesium metal as a vapor at the high temperatures (step 3). Although the formation of SiO2 is thermodynamically favored over SiO [18], it is likely that under many processing conditions, the availability of atomic oxygen at the interface is insufficient to account for the total formation of SiO2. This would require the SiO to scavenge oxygen from the MgO film, per step 2. This secondary oxygen scavenging step can account for the experimentally observed
3196
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
increase in interface breakdown between the two plasma exposures, where the harsh plasma (i.e. more available atomic oxygen) showed greater interface breakdown and increased loss of MgO film thickness. From Figs. 2 and 3, the increase in interface breakdown is influenced by both the availability of atomic oxygen and substrate temperature. Fig. 3 indicates a kinetically controlled mechanism where the availability of atomic oxygen at the interface is in excess, and the rate of interface breakdown is kinetically controlled by substrate temperature. Fig. 2, which compares the formation of SiOx at 650 8C to increasing atomic oxygen in the vacuum environment, indicates that the rate of interface breakdown is influenced by the availability of atomic oxygen at the interface. Although the observed contributions of both temperature and atomic oxygen support a kinetically controlled mechanism, it is not possible to rule out the influence of the diffusion rate for atomic oxygen along MgO grain boundaries. The MgO film heated in vacuum (sample 1-post) remained chemically stable based on XPS characterization of the Mg2p photoelectron peak before and after each exposure. In addition, RHEED indicated a slight 3D to 2D transition after the substrate temperature reached 650 8C. When the heated sample was cooled to room temperature, the more 2D RHEED pattern was maintained. A typical RHEED pattern for a MgO film deposited at 140 8C is represented in Fig. 4b. The RHEED pattern is characteristic of MgO(1 1 1) taken along the < 1 1¯ 0 > azimuth and shows some 3D characteristics, as indicated by the intensity spots located in the zeroth Laue zone. During the temperature exposure step of sample 1 to 740 8C in vacuum, the 3D RHEED characteristics were observed to transition to 2D characteristics when the substrate temperature reached 650 8C. The film maintained the (1 1 1) orientation throughout the high-temperature exposure. The transition to 2D characteristics is indicated by the smoother streaks in the zeroth Laue zone, represented in Fig. 4c. This apparent RHEED transition at 650 8C was observed for all four samples that were exposed to different temperature and oxygen conditions. To ˚ MgO(1 1 1) was further explore this transition, a 20 A deposited on 6H-SiC(0 0 0 1) at 140 8C followed by rapidly heating the sample to 650 8C at 1.0 109 Torr. AFM ˚ MgO(1 1 1) film characterization was performed on the 20 A ˚ as-deposited after heating, and was compared to another 20 A film. Based on the 3D characteristics observed by RHEED, it
was expected that the as-deposited film would show higher surface roughness compared to the heated sample. The average surface roughness from five locations on the as-deposited MgO(1 1 1) and the heated MgO(1 1 1) was measured to be 0.27 0.02 nm and 0.26 0.03 nm, respectively. Thus no measurable difference in surface roughness was observed within the resolution limitations of the AFM. No mention has been made up to this point regarding the termination of the MgO(1 1 1) surface. Due to the nature of the (1 1 1) orientation of ionic rocksalt structures, including MgO, the alternating layers of oppositely charged magnesium and oxygen result in a dipole moment perpendicular to the surface. In order to minimize this instability, the surface cannot truly be terminated by a layer of oxygen or magnesium. Much research has been conducted to understand the true surface termination of MgO(1 1 1) [23–25,29]. Early work suggested the surface became faceted into neutral {1 0 0} planes [23,24]. However, more recent research has reported that faceting is not a favorable means of stabilization, but rather MgO(1 1 1) surface terminations of H3 H3 R308, 2H3 2H3 R308, or 2 2 reconstructions are predicted to be more favorable for minimizing electric dipoles [25–29]. Although these surface reconstructions are electrically stable, they often require high temperatures (>800 8C) to form. Further research combining experimental and theoretical studies supported a hypothesis for an OH terminated MgO(1 1 1)-(1 1) surface, which is most energetically favorable and results in a surface termination that is more stable than the non-polar MgO(1 0 0) surface [29–31]. Through careful characterization of our MgO films by XPS ˚ and RHEED we see evidence of this OH termination. A 100 A MgO(1 1 1) film was deposited on hydrogen cleaned 6HSiC(0 0 0 1) at a substrate temperature of 140 8C. A thickness of ˚ was chosen to ensure that the oxygen O1s photoelectron 100 A signal was originating from only the MgO film and not the ˚ is greater than the silicate adlayer on the 6H-SiC, since 100 A attenuation depth of electrons originating from oxygen at the interface. The resulting RHEED pattern indicated some 3D characteristics, as expected, but maintained a pattern characteristic of MgO(1 1 1). XPS characterization of the film indicated one Mg2p photoelectron peak, but two O1s photoelectron peaks. The primary O1s peak, associated with MgO, was located at a binding energy around 530 eV. The secondary peak was located at a binding energy 2 eV higher and was designated as the OH surface state, which is in close agreement with Lazarov et al. [31].
Fig. 4. RHEED patterns for MgO growth and heat exposure illustrating the heat-induced transition from 3D characteristics to 2D characteristics. (a) Hydrogen ˚ MgO(1 1 1) deposited on hydrogen cleaned 6H-SiC(0 0 0 1) at 140 8C, and (c) 20 A ˚ MgO(1 1 1) film cleaned 6H-SiC(0 0 0 1), (b) heteroepitaxial growth of 20 A annealed rapidly to 650 8C at 1.0 109 Torr and held for 2 min.
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
Taking into account sampling volume, the relative intensity of the OH peak to the MgO peak was calculated to be 0.11, indicating slightly more than one monolayer coverage. Quantitative characterization of the actual monolayer coverage would require the use of angle resolved XPS. However, this was not available, leaving the relative ratio of OH to MgO as a qualitative measure. The monolayer coverage of OH greater than 1 is most likely due to the 3D features observed by RHEED, which would increase the overall surface area and therefore increase the amount of available sites for OH termination. ˚ MgO film was grown at 140 8C and was then A 100 A annealed to 650 8C at 1.0 109 Torr for 2 min in order to transition to the more 2D surface, as observed through RHEED. The more 2D RHEED pattern was observed on the 650 8C film as well as when the film had cooled to room temperature. XPS characterization of the film indicated a significant drop in the relative amount of OH to MgO. On the annealed film, the ratio of OH to MgO, as indicated by the O1s peak, was calculated to be 0.08. The decrease in the relative amount of OH is most likely due to a surface smoothing effect. Additionally, a third sample was deposited to a thickness of ˚ at a 650 8C growth temperature in order to maintain a 100 A smooth 2D surface during growth and then to evaluate the surface termination after cooling. XPS characterization of the film indicated a relative OH to MgO ratio of 0.06 as calculated from the O1s peak. The additional decrease in the relative ratio can again be associated with the smoother 2D surface that was maintained during film growth. Fig. 5 illustrates the O1s ˚ low-temperature MgO, photoelectron spectra for the 100 A ˚ annealed MgO, and 100 A high-temperature MgO. XPS characterization of all the films discussed indicate a MgO(1 1 1)-(1 1) OH terminated film. The decrease in OH to MgO is most likely associated with a smoother MgO surface, which was best observed by RHEED.
˚ MgO(1 1 1). Bottom spectra is for MgO Fig. 5. XPS spectra of O1s for 100 A as-deposited at 140 8C, middle spectra is the same MgO film annealed to 650 8C for 2 min, and top spectra is Mg-O deposited at 650 8C. The decrease in Mg-OH is believed to be related to the observed RHEED transition from 3D to 2D, which results in a smoother surface.
3197
The smoother MgO surface resulted in less 3D features and therefore less surface area to be terminated by the OH group. It is currently unclear if the decrease in 3D features is related to a decrease in grain boundaries or only surface smoothing. If the MgO film deposited at 650 8C does in fact have fewer grain boundaries, then this may help better understand the contribution of oxygen diffusion along grain boundaries, which is being considered for future in-depth understanding of the interface breakdown mechanism. Although the MgO film deposited on the 6H-SiC at 650 8C was too thick to observe any interface breakdown by XPS, similar high temperature growths have ˚ . Interestbeen performed with MgO thicknesses around 20 A ingly, although the substrate was shuttered from the line-of-site of the oxygen source, even at substrate temperature of 650 8C and an oxygen environment of 5.0 106 Torr and 35 mV there was no evidence by XPS or RHEED of surface oxidation prior to growth. This may be due to the thermodynamically favored formation of MgO compared to SiO. Although SiO2 is thermodynamically favorable over both SiO and MgO, it does require bonds to be broken within the silicate adlayer of the SiC. Therefore, it is not too surprising that the formation of MgO is favorable given the availability of atomic magnesium and atomic oxygen at the surface. Degradation of the MgO/SiC interface may be detrimental to ultimate device performance. In order to maintain an effective MgO/SiC interface, the subsequent deposition of the functional oxide should avoid long-term exposure of the MgO layer to temperatures above 650 8C in an oxygen environment, especially an active oxygen environment such as the oxygen plasmas often used in MBE depositions. To test the feasibility and effectiveness of MgO(1 1 1) as a template layer for functional oxides, BTO was deposited by MBE on MgO/SiC and bare SiC. When deposited on the bare 6H-SiC, the resulting BTO was amorphous. Although no interface breakdown or oxidation of silicon was observed, the H3 H3 R308 surface of the 6H-SiC(0 0 0 1) did not promote the heteroepitaxy of BTO. Based on the crystal structure and lattice spacing of the 6H-SiC surface reconstruction and the O–O spacing of the BTO(1 1 1) there are two likely alignments. One has a compressive mismatch of 6.2% with a 308 in-plane rotation of the BTO(1 1 1) on the 6H-SiC silicate adlayer. The second has a tensile mismatch of 7.9% with no inplane rotation, but does require two atomic spacings of BTO(1 1 1) to one reconstructed 6H-SiC. Fig. 6a and b show the RHEED patters for the BTO films deposited on bare 6H-SiC. ˚ MgO template layer under When BTO was deposited on the 20 A the same growth conditions, the resulting BTO was crystalline with a distinct, transmission dominated RHEED pattern characteristic of a twinned (1 1 1) oriented tetragonal perovskite structure (Fig. 6e). XRD of the BTO film resulted in only (1 1 1) orientated BTO. The observed heteroepitaxy of the BTO on the MgO template layer is most likely do to the improved lattice matching. The calculated lattice mismatch based on the observed epitaxial relationship and the O–O spacing for both MgO(1 1 1) and BTO(1 1 1) is 4.4% in tension. The decrease in tensile mismatch from 7.9%, for BTO/6H-SiC, to 4.4%, for BTO/MgO/ 6H-SiC promoted the heteroepitaxial integration of BTO on 6HSiC. This demonstrates the importance and effectiveness of
3198
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199
Fig. 6. RHEED patterns illustrating the importance of the MgO(1 1 1) template layer for integrating crystalline BTO on 6H-SiC(0 0 0 1). The crystalline MgO(1 1 1) layers proved essential for the heteroepitaxy of the BTO film.
crystalline MgO(1 1 1) as a heteroepitaxial template layer for integrating a wide variety of (1 1 1) oriented perovskite ferroelectrics on the wide band gap semiconductor 6H-SiC.
thank Dr. Al Sacco at Northeastern University for use of the AFM, and Lena Fitting Kourkoutis at Cornell University for independent confirmation of film thickness by TEM.
4. Summary and conclusions References The post-deposition effects of heating and oxygen species on chemistry, morphology, and crystallography of MgO(1 1 1) on ˚ ) MgO films were 6H-SiC(0 0 0 1) were evaluated. Thin (20 A found to be thermally stable when heated to 740 8C at 1.0 109 Torr, and when exposed to a harsh plasma pressure of 5.0 106 Torr at 140 8C. However, interfacial breakdown and surface roughening began to occur when the films were exposed to molecular oxygen or oxygen plasma at 650 8C for 60 min. Exposure to a constant harsh plasma at different temperatures indicated a kinetic component to the interface breakdown where the rate of interface breakdown was found to increase significantly at temperatures above 400 8C. Also occurring at 650 8C was a transition from a 3D to more 2D RHEED pattern and a reduction in OH surface termination in the MgO film, most likely due to the decrease in surface area of the smoother MgO. Despite these changes, RHEED patterns indicated the preservation of the MgO(1 1 1) orientation under all conditions. Successful deposition of crystalline BTO(1 1 1) was obtained by MBE when deposited on the MgO/6H-SiC, but not on the hydrogen cleaned 6H-SiC. The improved epitaxy demonstrates the potential for using crystalline MgO as a heteroepitaxial template layer for integrating a wide range of perovskite ferroelectrics on 6H-SiC. Acknowledgements This research was supported by Office of Naval Research under contracts N0014-06-1-0761 and the N00014-04-1-0426, Dr. Colin Wood Program Monitor. The authors would like to
[1] W.A. Doolittle, A.G. Carver, W. Henderson, J. Vac. Sci. Technol. B 23 (2005) 1272. [2] Z. Chen, A. Yang, S.D. Yoon, K. Ziemer, C. Vittoria, V.G. Harris, J. Magn. Magn. Mater. 301 (2006) 166. [3] A.R. Meier, F. Niu, B.W. Wessels, J. Cryst. Growth 294 (2006) 401. [4] T.L. Goodrich, J. Parisi, Z. Cai, K.S. Ziemer, Appl. Phys. Lett. 90 (2007) 042910. [5] T.L. Goodrich, Z. Cai, M.D. Losego, J.-P. Maria, K.S. Ziemer, J. Vac. Sci. Technol. B 25 (2007) 1033. [6] J.-J. Chen, B.P. Gila, M. Hlad, A. Gerger, F. Ren, C.R. Abernathy, S.J. Pearton, Appl. Phys Lett. 88 (2006) 042113. [7] H.S. Craft, J.F. Ihlefeld, M.D. Losego, R. Collazo, Z. Sitar, J.-P. Maria, Appl. Phys. Lett. 88 (2006) 212906. [8] F. Niu, B.H. Hoerman, B.W. Wessels, J. Vac. Sci. Technol. B 18 (2000) 2146. [9] J. Zhu, X.H. Wei, Y. Zhang, Y.R. Li, J. Appl. Phys. 100 (2006) 104106. [10] J.H. Haeni, et al. Lett. Nat. 430 (2004) 758. [11] K.A. Shaw, E. Lochner, D.M. Lind, J. Appl. Phys. 87 (2000) 1727. [12] J. Bernhardt, J. Schardt, U. Starke, K. Heinz, App. Phys. Lett. 74 (1999) 1084. [13] T. Goodrich, J. Parisi, J. Leong, K.S. S Ziemer, AIChE Annual Meeting Conference Proceedings, Cincinnati, OH, United States, Oct. 30–Nov. 4, (2005), p. 135e/1. [14] C.D. Theis, D.G. Schlom, J. Vac. Sci. Technol. A 14 (1996) 2677. [15] NIST Database 82, U.S. Department of Commerce (2001). [16] Y.R. Li, Z. Liang, Y. Zhang, J. Zhu, S.W. Jiang, X.H. Wei, Thin Solid Films 489 (2005) 245. [17] http://srdata.nist.gov/xps. [18] M.W. Chase, Jr., C.A. Davies, J.R. Downey, Jr., D.J. Frurip, R.A. McDonald, A.N. Syverud, J. Phys. Chem. Ref. Data 14 (1985) JANAF Thermochemical Tables. [19] X.Y. Chen, K.H. Wong, C.L. Mak, J.M. Liu, X.B. Yin, M. Wang, Z.G. Liu, Appl. Phys. A 74 (2002) 703. [20] D. Schmeiber, D.R. Batchelor, R.P. Mikalo, P. Hoffmann, A. Lloyd-Spetz, Appl. Surf. Sci. 184 (2001) 340.
T.L. Goodrich et al. / Applied Surface Science 254 (2008) 3191–3199 [21] G.G. Jernigan, R.E. Stahlbush, N.S. Saks, Appl. Phys. Lett. 77 (2000) 1437. [22] W. Chen, X.N. Xie, H. Xu, A.T.S. Wee, K.P. Loh, J. Phys. Chem. B 107 (2003) 11597. [23] V.E. Henrich, Surf. Sci. 57 (1976) 385. [24] H. Onishi, C. Egawa, T. Aruga, Y. Iwasawa, Surf. Sci. 191 (1987) 479. [25] R. Plass, K. Egan, C. Collazo-Davila, D. Grozea, E. Landree, L.D. Marks, M. Gajdardziska-Josifovska, Phys. Rev. Lett. 81 (1998) 4891. [26] D. Wolf, Phys. Rev. Lett. 68 (1992) 3315.
3199
[27] F. Finocchi, A. Barbier, J. Jupille, C. Noguera, Phys. Rev. Lett. 92 (2004) 136101. [28] A. Subramanian, L.D. Marks, O. Warschkow, D.E. Ellis, Phys. Rev. Lett. 92 (2004) 026101. [29] A. Wander, I.J. Bush, N.M. Harrison, Phys. Rev. B 68 (2003) 233405. [30] K. Refson, R.A. Wogelius, D.G. Fraser, M.C. Payne, M.H. Lee, V. Milman, Phys. Rev. B 52 (1995) 10823. [31] V.K. Lazarov, R. Plass, H.-C. Poon, D.K. Saldin, M. Weinert, S.A. Chamber, M. Gajdardziska-Josifovska, Phys. Rev. B 71 (2005) 115434.