Amiya Banerjee*,
K. V. L. V. Narayanachari and
Srinivasan Raghavan
Centre for Nano Science and Engineering, Indian Institute of Science, India. E-mail: amiya.matsci@gmail.com
First published on 23rd March 2017
Yttria stabilized zirconia (YSZ) films are being used both as functional oxide and buffer layers for integration of various other functional oxide films on Si substrates. As functional properties of these oxides are highly anisotropic in nature, highly oriented films are essential to realizing their fascinating properties. (111) and (100) textured YSZ films have been deposited on Si substrates by reactive-direct current (R-DC) sputtering. Annealing of these films leads to grain growth and improvement in texture. However, it strongly depends on the growth stresses developed during deposition of these films. Depending on stress harnessing in films/stacks, the texture was improved from rocking curve FWHM of 16° to 7° and 25° to 15° for (111) and (100) YSZ films respectively. A detailed analysis of the relation between stress and grain growth is carried out using an energy balance model. We have found that grain growth is limited by kinetics, though it should be possible from a thermodynamic viewpoint. It is observed that higher initial compressive stress aids significant grain growth (∼150%) and texture-improvement (∼57%) on annealing.
Due to its difference in chemical nature, growth of oxides on Si is more challenging than that on oxide substrates. Interfacial energies can be large due to both strain and differences in the chemical nature of bonding. In addition, the Si surface is invariably covered with an amorphous oxide layer that is stable in the oxygen ambient required for the growth of most oxides themselves.13,14 Hence, establishing epitaxy to obtain high quality crystalline oxide films, as on oxide substrates, is a challenge. As a result, the films grown are strained, defective and polycrystalline with no preferred orientation. Thus, when they are deposited in the form of thin films on Si substrates, these oxides typically do not show properties as good as those of their bulk single crystals. If not epitaxial, films with fibre texture-polycrystalline films in which one crystallographic direction is parallel to the normal to the film substrate interface, but there is in-plane rotation around this direction-can also be enable many applications that exploit anisotropic functional properties. The films to be discussed in this paper all have fibre and are not epitaxial.
In spite of similar challenges and growth mode (Volmer–Weber), another analogous system, GaN on Si, has been successfully commercialized by metal organic chemical vapour deposition (MOCVD). The ability to reduce starting defect densities by understanding the correlation between stress and defect evolution during growth played a key role in enabling the integration of these nitrides with Si for a wide variety of applications from light emitting diodes to high power and high frequency transistors.15 However, for oxides there is very little to no information in the literature on the interplay between defects and stress, intrinsic and extrinsic, which could potentially be exploited to achieve such an improvement in properties. Hence, a study on the relationship between stress and defects in oxide films is necessary.
Our previous studies indicated that even on Si surfaces with native oxide layer, control of crystalline orientation and stress can be achieved in films of yttria stabilized zirconia (YSZ) deposited by sputtering.16 YSZ is stable against the formation of native oxide of silicon17 and hence had been chosen as a buffer layer for the integration of films of different oxides.18,19 It is very important to note that most of the properties of oxides are anisotropic in nature.20,21 As a result, highly oriented polycrystalline films exhibit better physical parameters than their randomly oriented counterpart. Therefore, oriented oxide films are desired for many applications. As crystallinity and orientation of functional oxide films are strongly dependent on the orientation of the buffer layer, it is beneficial to have a highly oriented buffer layer. Oriented or textured YSZ buffer layers are achieved either by pulsed laser deposition, or by modified sputtering such as ultra-high vacuum ion sputtering and metallic target sputtering.18,19,22–26 However, the sputtered YSZ films are in general polycrystalline and there are few reports of textured YSZ films on Si by sputtering.16,27–31 Additionally, there are very few reports to the best of our knowledge on the quality of textured YSZ films on Si as determined by using rocking curve. It is important to mention that FWHM of rocking curve (ω-scan) is a measure of the quality of a textured film. Lower the value of FWHM of rocking curve, better is the quality. This study shows that it is possible to harness stress to bring about a significant reduction in defect levels in oxide films compared to those formed during the initial stages of growth. The observation agrees well with the grain growth–stress evolution model proposed in literature.32
We chose sputtering as a deposition method as it is inexpensive and routinely used in Si CMOS fabrication for deposition films over larger areas. While, reports on (111) and (100) textured YSZ films on Si substrates by controlling various growth parameters such as growth rate,16 growth temperature,30 O2/Ar ratio31 and substrate bias33 exist, very few reports are available on the stress–texture evolution correlation. This effect is analysed here based on an energy balance model.32 Details of the analysis and improvement in the texture of YSZ films are presented in the result and discussion section. Understanding the correlation among stress–grain growth–texture evolution enabled us to achieve the best quality sputtered (111) YSZ films with a FWHM of rocking curve 7°. Films with FWHM as low as 1° have been obtained by pulsed laser deposition.18,34 However, in comparison to sputtering, it is more difficult to deposit uniform thin films over large areas by PLD. Though the effect of stress on grain growth has been studied for metals, this is the first report on correlation between stress, grain-growth and texture-evolution in oxide film. The results obtained in this study can also be applied to other oxide films for enhancing crystalline quality and hence corresponding properties.
An X-ray diffraction (θ–2θ) scan of a film shows relative density of various grain orientations in the film. As can be noticed in the Fig. 2(a)–(c) that only one peak of YSZ in each plot is present (peaks seen with a star mark belong to the underlying Si) indicating that the YSZ films are textured or oriented. YSZ films deposited at lower rate (0.8 nm min−1) are (111) oriented [Fig. 2(a)], whereas those at higher growth rate (30 nm min−1) are (100) oriented [Fig. 2(b)]. The third pattern [Fig. 2(c)] corresponds to a film that was nucleated and grown at high growth rate (20 nm min−1) for the (100) texture to develop (refer to Narayanachari et al.16 for more details). Following this initial growth, deposition was continued at a slower rate (0.4 nm min−1). As can be noted, the texture does not change and remains as (100). The reason to grow this stack will be apparent later in this paper.
As mentioned earlier the aim was to study the effect of film stress on defect densities of YSZ films and hence it is necessary to control stress. In order to have better control over stress, it needs to be monitored during deposition.35 In situ stress measurements of the as-grown (111) and (100) textured YSZ films on Si are as shown in Fig. 3 and 4 respectively. At low rates, films grow in compression, due to adatom injection into grain boundaries, and at higher rates in tension, due to boundary formation, as explained in ref. 16 in great detail. The slope of stress-thickness versus thickness plot is a measure of incremental in-plane stress in a film. Positive slope indicates tensile stress, whereas negative slope indicates compressive stress in a film36 (also see schematic in ESI S-2†). It can be seen in Fig. 3 that with a decrease in the growth rate there is an increase in the magnitude of compressive stress in the (111) oriented films.
Fig. 3 Stress thickness versus thickness plots show the growth stresses for (111) YSZ films deposited at different growth rates. |
In Fig. 4(a) and (b) the two stress thickness vs. thickness plots of the (100) oriented films correspond to the two XRD patterns shown in Fig. 2(b) and (c) respectively. It can be noticed that a film deposited at high growth rate (30 nm min−1) grows completely under a tensile stress [Fig. 4(a)]. A stack, which was initially deposited at high growth rate (20 nm min−1) for the (100) texture to develop and followed by lower rate (0.4 nm min−1), exhibited a change in growth stress from tensile to compressive [see Fig. 4(b)]. The reason for the deposition of this stack will become clear in the sections to come.
A summary of all the (111) and (100) YSZ films deposited at various deposition rates and having different stresses are presented in the Fig. 5 to give a clear picture in readers' mind.
The θ–2θ scans indicate that the as-deposited YSZ films are textured in Fig. 2. However, Fig. 6 and 7, ω or rocking curve scans, show that both the as-deposited (111) and (100) films are highly defective as represented by their large FWHM of 16° and 25° respectively.
A summary of the variation in FWHM with growth stress (σg) of the YSZ films is shown in Fig. 8(b). It shows that the FWHM of the rocking curve decrease from 25° to 16° with growth rates ranging from 30 to 0.8 nm min−1 for (100) and (111) YSZ films respectively. This is due to compressive stress driven annealing during growth which will discussed in detail as we go along. Thus, (100) films are more defective than (111) films as (100) YSZ films show larger FWHM than (111) films. A measure of how defective these oxide films are can be gauged from the fact that the as-deposited rocking curve widths of epitaxial III-nitride films on Si are routinely in the 0.1–1° range.
In order to improve crystallinity, the deposited oxide films were annealed at 1000 °C for one hour under ambient condition. The improvement in texture is quite substantial for films with compressive stresses as shown in Fig. 6 and 8. In particular, for the film with a compressive stress of −0.8 GPa, the FWHM was found to reduce by 50% from 16.3 to 8.2. This sharp reduction is accompanied by a 250% increase in the grain size from 20 nm to 70 nm as measured by the AFM scans shown in Fig. 9 (top), also by a 150% increase in vertical coherence length from 10 nm to 25 nm, calculated from X-ray diffraction peak broadening (see ESI S-4, Fig. S3†). In contrast, for tensile stressed (100) YSZ film, the increase in AFM grain size is only 30% [Fig. 9 (bottom)], and 16% as measured from X-ray diffraction peak broadening (see ESI S-4, Fig. S6†). It is hence conclusive that compressive stresses lead to considerable grain-growth of YSZ films. Similar compressive stress-driven grain growth have been observed for metallic films.37,38 Grain growth leads to a reduction in FWHM. The improvement in FWHM of rocking curves implies that larger energy high angle grain boundaries are replaced by lower energy small angle grain boundaries during such growth. Therefore, the reduction in FWHM for the (100) YSZ film is also only 25° to 20°. It can be concluded from these two results that improvement in texture for the compressively stressed film is more compared to the tensile stressed one. It is also observed that improvement in crystallinity during annealing is more with increase in initial compressive stress in these films (see Fig. 6). The lowest FWHM of 7° was achieved in heavily compressively stressed film (see Fig. 6), which is better in terms of crystalline quality than any sputtered (111) YSZ film reported before. In the tensile stressed (100) film FWHM was 20° after annealing. It was observed that compressive stress helped in significant texture improvement for (111) YSZ films. Therefore, to improve the texture of (100) YSZ films further, a compressive layer (∼125 nm) was deposited on top of a (100) tensile layer (∼275 nm) [see schematic in Fig. 4(b)]. The FWHM of this two layered stack after annealing is found to be furthered reduced to 14° [see Fig. 7(c)] which indicates a remarkable improvement in texture. Assuming that the bottom layer has a FWHM of ∼20° as measured for other (100) film [in Fig. 7(b)], at the very least the compressive stress has resulted in a dramatic drop to 14° of the whole stack. The actual FWHM from the top layer alone is expected to be much lower. Thus, even in this case, the compressive top layer has helped significantly to enhance texture of the whole stack.
Our observations indicate that compressive stress in as-deposited film helps to improve texture on annealing. This behaviour is now analysed by using Chaudhari's model32 for grain growth upon annealing. In general, the driving force for the grain growth during annealing is the reduction in excess free energy associated with grain boundaries. The densification associated with the reduction in grain boundary area per unit volume however results in shrinkage. For a thin film constrained by a substrate, such shrinkage results into in-plane tensile stress in the film, thereby leading to an increase in the overall mechanical energy of the film. A starting in-plane compressive stress is thus expected to be useful as has already been observed. Grain growth proceeds when the overall change in the energy (ΔEFilm) becomes negative for the process and the ΔEFilm is be expressed as
ΔEFilm = ΔEStrain + ΔEBoundary |
Fig. 10 Critical grain size vs. initial stress plot shows dependence of critical grain size on initial in-plane stress in a film. |
In the former condition (for d0 < dcr), where there is an energy minimum, grain growth on annealing would be arrested once a system reached the minimum. While in the latter (for d0 > dcr), annealing would result in the continued grain growth and extended generation of a tensile stresses.
In order to apply the model, we use the crystallite size as measured by X-ray line broadening in θ–2θ scans (all the θ–2θ scans of the as-deposited and annealed YSZ films are provided in the ESI S-4†). In doing so, the implicit assumption being made, see ESI S-5,† is that grain size and growth is equiaxed. The starting stress (σo) value used in these calculations is the sum of the growth stress (σg) measured and thermal mismatch stress generated during the heating from growth temperature (700 °C) to the annealing temperature (1000 °C). In the present case the thermal mismatch stress is −0.5 GPa (see ESI S-3†).
Fig. 11 shows the change in film energy vs. final grain size of (111) YSZ films with d0 < dcr. It can be observed that there is an energy minimum for each film. Final grain sizes of these films after annealing matches with the corresponding energy minimum in top of Fig. 11. There is a close match between the experimental values (Fig. 11 top) and the calculated value the model (Fig. 11 bottom). It can also be noticed that higher the initial compressive stress, larger the final grain size.
Fig. 12 shows plot of change in film energy vs. final grain size for (100) YSZ films with d0 > dcr. It can be noticed that there is no energy minimum as expected from the energy balance model. However, experimental results in the corresponding top figure shows grain growth saturation at some particular grain size. Note that the x-axis is up to 60 nm to make grain sizes clear, but no energy minimum was found even when x-axis was extended to >150 nm. This contradicts the model which predicts that if d0 > dcr, grain should grow continuously and grain growth should not get saturated.
To confirm whether the grain growth is actually limited by kinetics and not by thermodynamics, two of these films with d0 > dcr were annealed at a higher temperature (1200 °C) for another hour. If there is no grain growth even at this higher temperature, then it can be concluded that the grain growth saturates due to limitation in diffusion of grain boundaries. It was observed that there was hardly any increase in grain size (see Fig. 13). Thus, the obvious reason for the lack of agreement with thermodynamic predictions seems to be due to the kinetics factors. It is to be noticed that the slope of the energy-grain size plot (J m−1 or ) can be looked upon as a measure of kinetic driving force for grain growth. This slope for a film with d0 < dcr is higher (∼one order of magnitude) than for a film with d0 > dcr (see Fig. 11 and 12). Therefore, kinetic driving force for atomic diffusion aiding grain growth is smaller for the latter case. Hence, grain growth was found to be restricted by kinetics, though thermodynamically there was no restriction for those films with d0 > dcr. While a detailed modelling of grain growth is out of the scope of this paper, it is very well known that the rate of grain growth scales inversely with the grain size.40 This is another reason which supports the experimental observation that there is kinetic limitation for grain growth in (100) YSZ films with d0 > dcr. Thus, while grain growth in a film with large initial grain size (also d0 > dcr) might not be restrained by a thermodynamic minimum, it appears that it will be limited by kinetic factors.
Fig. 13 Time of annealing versus grain size plots show grain growths for films with different growth stresses. |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra28437j |
This journal is © The Royal Society of Chemistry 2017 |