Understanding the microstructural evolution and mechanical properties of transparent Al-O-N and Al-Si-O-N films

ABSTRACT Optically transparent, colorless Al-O-N and Al-Si-O-N coatings with discretely varied O and Si contents were fabricated by reactive direct current magnetron sputtering (R-DCMS) from elemental Al and Si targets and O2 and N2 reactive gases. The Si/Al content was adjusted through the electrical power on the Si and Al targets, while the O/N content was controlled through the O2 flow piped to the substrate in addition to the N2 flow at the targets. The structure and morphology of the coatings were studied by X-ray diffraction (XRD) and transmission electron microscopy (TEM), while the elemental composition was obtained from Rutherford backscattering spectrometry (RBS) and heavy ion elastic recoil detection analysis (ERDA). The chemical states of the elements in the coatings were analyzed by X-ray photoelectron spectroscopy (XPS). Based on analytical results, a model describing the microstructural evolution of the Al-O-N and also previously studied Al-Si-N [1, 2, 3, 4] coatings with O and Si content, respectively, is established. The universality of the microstructural evolution of these coatings with the concentration of the added element is attributed to the extra valence electron (e–) that must be incorporated into the AlN wurtzite host lattice. In the case of Al-O-N, this additional valence charge arises from the e – acceptor O replacing N in the AlN wurtzite lattice, while the e – donor Si substituting Al fulfills that role in the Al-Si-N system. In view of future applications of ternary Al-O-N and quaternary Al-Si-O-N transparent protective coatings, their mechanical properties such as residual stress (σ), hardness (HD) and Young’s modulus (E) were obtained from the curvature of films deposited onto thin substrates and by nanoindentation, respectively. Moderate compressive stress levels between −0.2 and −0.5 GPa, which suppress crack formation and film-substrate delamination, could be obtained together with HD values around 25 GPa.


TSTA_A_1666425
Information Classification: General and Bragg-Brentano geometry. All measurements apart from RCs were performed with the detector in 1D mode, while RCs were recorded in 0D mode. Data were analyzed with proprietary software and with Matlab (The MathWorks, Inc.).
Symmetric  -2 scans were recorded to obtain information about crystal planes parallel to the sample surface and thus orthogonal to the film growth direction (z). The obtained line profiles (LPs) are shown in figs. 9 and 10 and were analyzed via line profile analysis (LPA) described in the book of Birkholz [11], pages 85-101. A pseudo-Voigt (pV) function was used for the fitting of each wurtzite (002) peak, separating Cu K 1  and K 2  . The peak position of the K 1  component was converted into the c-axis lattice spacing of the unit cell. The peak profile of the K 1  component was deconvoluted into its Cauchy and Gauss contributions. The Cauchy contribution to the peak broadening, assessed through the integral breadth, yielded the crystallite size (CS) via Scherrer's equation and the Gauss contribution the microstrain (MS) in the measured film. The instrumental broadening was accounted for via Si(400) peak obtained from the wafer substrate. RCs on the (002) peaks were analyzed for their FWHM, shown in fig.  11, as a measure for the angular distribution of crystallite tilts away from z. For the RCs, 2 was fixed at the measured Cu K 12  + maximum of the wurtzite (002) peaks, and  was rocked from 0  -2 .
Pole figures (PFs) of wurtzite (002), (103) and (101), shown in fig. 12, were taken to explore the 3D texture of coatings. The "thinned" PF mesh function of the software was used with a delta of 2  , giving 5000 measurement spots distributed on 40  circles equidistantly spaced on the 0-80   radius. For the (002) PF of a film, the 2 value effectively measured in the symmetric  -2 scan of the respective film was used. For (103) and (101)  During high temperature in situ XRD (HTisXRD) up to 900  C, the heating stage was constantly flushed by Ar. After each heating step of 50  C, the sample was equilibrated at the corresponding temperature for 1 h and measured with a symmetric  -2 scan. A further scan was taken after re-cooling of the sample.
Height variations due to thermal expansion of the sample holder were corrected by a stage mover.
Transmission electron microscopy (TEM) to image the cross-sectional structure of Al-O-N coatings was carried out on a JEM-2200FS (Jeol Germany). Films on Si(100) were broken, shaped into a wedge by mechanical tripod polishing, and finally ion-milled to e − transparency. The wedge was contacted to a Cu ring with an inner diameter of 2 mm and then imaged in regions of 20-30 nm thickness. Bright field (BF) and dark field (DF) images, electron diffraction (ED) patterns and high resolution (HR) images were recorded.

TSTA_A_1666425
Information Classification: General

X-ray photoelectron spectroscopy (XPS) was performed on a PHI Quantum 2000 Scanning
ESCA Microprobe System (Physical Electronics (PHI)). Monochromatic Al K  radiation of 1486.6 eV was used with a beam diameter of 150.0  m. The base pressure in the measurement chamber was below 8 9 10 −  mbar. The energy scale of the spectrometer was calibrated with the binding energy (BE) of the photoelectron (photo e − ) lines Cu 2p 3/ 2 , Ag 3d 5/ 2 and Au 4f 7/ 2 at 932.62, 368.21 and 83.96 eV, respectively, to within  0.1 eV [ISO 15472;2010-05]. The FWHM of Au 4f 7/ 2 amounted to 0.72, 0.93 and 1.57 eV for pass energies of 5.85, 29.35 and 117.4 eV, respectively.
Prior to measurements, the films were sputter-cleaned by an Ar ion bombardment at 1 kV during 600 s to remove the oxide natively forming upon atmosphere exposure and eventual carbonaceous contaminations.
During measurements, the insulating samples were charge neutralized by low energy Ar ions and e − . To check for and, if necessary, correct for a potential charge buildup despite charge neutralization, an established Au referencing method [13] was applied. For this, the molybdenum (Mo) masks used to clamp the samples during XPS measurements had been coated with 100 nm Au beforehand, so that a small amount of Au was re-deposited on the sample surface during sputter cleaning for oxide removal.
Spectra of the photo e − lines Al 2p, Si 2p, O 1s and N 1s were measured at a pass energy of 29.35 eV and with 0.125 eV increments. For each peak, the BE and the FWHM were extracted. Spectral processing, shown exemplarily for an Al-Si-O-N film in fig. 13, was carried out in the software CasaXPS version 2.3.17PR1.1 (Casa Software Ltd.). Relative sensitivity factors were adopted from the system manufacturer PHI. Shirley backgrounds were applied. All peaks were symmetric and could be fitted with a single Gaussian-Lorentzian curve (GL(Gaussian percentage)). GL(30) for Al 2p and Si 2p, GL(40) for O 1s and GL(60) for N 1s were found to suit best. For N 1s, the second peak appearing at high BE for some samples was best fitted with GL(20).
Ellipsometry to assess the refractive index (n) and extinction coefficient (k) of coatings was measured on a spectroscopic ellipsometer M-2000 VI (J.A. Woollam). The linearly polarized beam with wavelengths in the range of 400-1700 nm was set to irradiate coatings on Si (100)  On substrates of conventional thickness, stress-induced cracks were observed in crystalline Al-O-N films of regime (I) (see section 5). Cross sections at locations of such stress-induced film cracks were prepared and imaged with a Gallium Focused Ion Beam SEM (GaFIB-SEM) FEI Helios NanoLab G3UC (FEI Company). A typical Ga-FIB-SEM image, showing a crack propagating along line of sight, is given in fig. 15. Nanoindentation for the hardness (HD) and Young's modulus (E) of the coatings was performed on a Hysitron UbI 1 (Hysitron, Inc.) equipped with a diamond Berkovich tip (SYNTON-MDP LTD.). Tip area calibrations on fused silica were regularly performed in between measurement sequences of 10-15 samples. Indentation load forces were chosen so that the resulting indent depths did not exceed 10% of the coating thickness in order to avoid an influence of the substrate on the measurement. 3 mN and/or 2 mN were found to be adequate and were used in quadratic 3x3 indentation matrices, in which the 9 single indents were spaced by 40  m. Depending on the thickness (800-1200 nm) and the HD (8-28 GPa) of the coatings, either both matrices or only the 2 mN matrix were taken. Typical indentation depths were in the range of 60-100 nm. For samples, for which the 3 mN matrix did not lead to too deep indents, no systematic difference between the 3 mN and the 2 mN matrix was found. The results were evaluated according to Oliver and Pharr [45].

Experimental data obtained for Al-Si-O-N
The microstructural evolution of quaternary Al-Si-O-N with the total Si and O content is found to be far more complex than that of the ternary systems. We attribute this to the formation of Si-O bonds which compete with the roles Si and O play for the microstructural evolution of the corresponding ternary coatings. No well-defined regime boundaries could be established for the quaternary system (see section section 6). This is shown by means of XRD data in fig. 16.
Four types of diffractograms are distinguished, and the compositions of the films yielding the line profiles representative for each type are given. Type I shows a dominant wurtzite (002) peak, which is right-shifted compared to that of pure AlN [43] This indicates that a Al-Si-O-N film contains crystallites with a shrunken c-axis wurtzite lattice parameter. Weak wurtzite (100) and (101) peaks also appear. Type (II) shows the right-shifted (002) peak with a tail extending towards lower 2 values. This tail possibly arises from a partially ordered grain boundary phase, or from a (100) component that is right-shifted due to a shrunken a-axis lattice parameter. In type III, the intensity of this tail component is increased. In type IV, a broad peak appears between 34 and 35  in 2 , while the intensity of the right-shifted (002) peak decreases.

Information Classification: General
There is a trend for films with higher (O+Si) content to yield diffractograms with a higher type number, but an unambiguous assignment of distinct chemical composition regimes to diffractogram types is not feasible. The film with the highest (Si+O) content showing crystalline diffraction signals contains 5.6% O and 17.4% Si and thus a total of 23%. Films with higher (Si+O) concentrations are X-ray amorphous.
As discussed in section 6, the residual stress  of the quaternary Al-Si-O-N coatings, shown in fig. 17, remains moderately compressive around -0.5 GPa with varying chemical composition.