Interface study on the effect of oxygen/nitrogen ratio in Ni/Ti multilayer deposited by reactive sputtering

Since the growth morphology along and perpendicular to the interface is important for supermirror applications, the dependence of this on the reactive gas has been investigated in Ni/Ti multilayers prepared by reactive magnetron sputtering with variable O2/N2 ratios. The interface properties are characterized by GIXRR, XDS, and TEM measurements. Compared to the case without O2, the presence of 20% O2 in the deposition of Ni layers contributes to smooth and abrupt interfaces. It also suppresses the accumulation of interfacial roughness with the increasing number of layers. However, the abundant oxygen content results in a striking degradation of interface quality associated with the crystallization evolution. Moreover, the lateral correlation length of interfacial roughness exhibits a consistent tendency with the grain size as the oxygen content increases. Following the XPS depth profiles, although N2 and O2 gases were applied in the Ni layer deposition, the N and O were only detected in the Ti layers as the compound for the high chemical activity of Ti. The elemental form in the Ni layers corresponds to the crystalline structure inferred by XRD measurements.


Introduction
Neutron supermirrors composing of Ni/Ti multilayers with the graded thickness can dramatically enhance the neutron flux and open the way for the effective transmission of thermal neutrons, which is crucial for the development of various neutron analyzing devices [1]. In the past decades, a lot of efforts have been devoted to improve the performance of Ni/Ti supermirrors such as using addition of C or N to Ni layers [2][3][4][5][6][7]. The addition of C or N reduces the grain size in Ni layers, which results in smoothing of the interfaces, leading to an improvement of the reflectivity in Ni/Ti supermirrors. However, the doping of individual element is effective to a limited degree. Scharpf et al observed that a stable layer structure of Ni/Ti supermirror can be realized by venting the vacuum after the deposition of each layer [8]. To achieve clean and controlled deposition, Kumar et al used synthetic air (about 80% N 2 and 20% O 2 ) instead of natural air as the reactive gas to prepare Ni/Ti supermirrors by reactive magnetron sputtering [9], which increases the reflectivity markedly compared to the Ni/Ti supermirrors prepared in pure Ar gas. Nevertheless, the above authors have mainly focused on the production technology of the supermirrors based on their own equipment. The effect of each dopant and the optimal proportion of dopants are key issues in the study of multielement doping. The previous studies were limited to a constant oxygen/nitrogen ratio. The effect of oxygen on the interface, crystallization, and chemical composition remains unclear. Up to now, the state-of-the-art reflectivity is much lower than the theoretical predictions of Ni/Ti supermirrors [10]. To obtain the higher reflectivity of Ni/Ti supermirrors, the microcosmic evolution behind the improvement of neutron-optics performance should be understood for further optimization. The study on the microcosmic properties of Ni/Ti multilayers deposited with the different oxygen/nitrogen ratios would be necessary both for the technology and the basic science.
The reflectivity of neutron supermirror depends on the constructive interference of reflective neutrons from each interface. Therefore, the morphology in the interface plane and along the growth direction play a crucial role in the performance [11][12][13]. In this paper, we studied the evolution of interfacial structures in Ni/Ti multilayers deposited with the variable O 2 /(N 2 +O 2 ) flow ratios using grazing incidence x-ray reflectivity (GIXRR), x-ray diffuse scattering (XDS), and transmission electron miscroscope (TEM), and then discussed the influence of microstructure and chemical composition on the interfacial structure using x-ray diffraction (XRD) and x-ray photoelectron spectroscopy (XPS).

Experimental details
Periodic Ni/Ti multilayers were deposited on Si (100) wafers at room temperature by reactive magnetron sputtering with a Ni target of 99.99% purity and a Ti target of 99.95% purity. The chamber was evacuated to a pressure of 1 × 10 −4 Pa prior to deposition. For the clean of the target surface contaminated by reactive gases, a pre-sputtering in pure Ar atmosphere was performed for more than 30 min. Each multilayer sample consists of 20 alternating Ni and Ti layers with the monolayer thickness of 10 nm. The Ti layers were prepared in pure Ar gas while the Ni layers were deposited reactively in Ar/N 2 /O 2 gas atmosphere. The flows of non-reactive and reactive gases were respectively fixed at 80 sccm and 20 sccm. The multilayer samples differ from the flow ratio of O 2 and N 2 defined as while R O 2 varied at 0, 20, 50 and 100%, respectively. To avoid the contamination of Ti layers by the reactive gases, after the deposition of each layer, the sample will stay at a waiting chamber for 1 min to allow time for extracting the residual reactive gases of experimental chamber.
GIXRR, XDS, and XRD measurements were conducted on a PANalytical Empyrean x-ray diffractometer supplied with 8.04 keV (λ = 0.154 nm) Cu Kα x-rays. The layer thickness and interfacial width including interfacial roughness and diffusion were determined by GIXRR measurements, performed in θ−2θ geometry over the range of 0°< 2θ < 6°. The details of interfacial roughness at various length scales were characterized by XDS, made in 'detector scan' geometry, i.e., the detector angel is varied while the incidence angle is fixed at firstorder Bragg peak. The lateral correlation length ξ || , vertical correlation length ξ ⊥ , and fractal exponent h of roughness structures were determined by fitting XDS data using the distorted wave Born approximation (DWBA). The crystalline properties of multilayers were investigated by XRD measurements, performed in grazing mode keeping the incidence angle fixed at 0.5°, to minimize the background from the diffraction of substrate Si lattice.
To observe intuitively the microstructure of the Ni/Ti multilayers deposited with different N 2 /O 2 ratios, the multilayer samples were thinned to be electron-transparent by a focused ion beam (FIB) with a gallium ion source at the working voltage of 30 kV. The obtained thinning samples were observed by the transmission electron miscroscope (TEM) operating at 200 kV.
Depth profiles of major elements of the Ni/Ti multilayers deposited with variable N 2 and O 2 gas conditions were performed using XPS measurements. The instrument used for the XPS spectra is Thermo Fisher Scientific Nexsa equipped with a monochromatized Al Kα x-ray source (1486.6 eV) operating at the power of 72 W. The core level spectra were obtained with 5 scans at 60 eV pass energy. Sputtering in the XPS measurements used Ar + with a raster area of 0.4 mm × 0.4 mm. The spectra were then deconvoluted using the Avantage XPS software. All the high-resolution peaks were fitted using the mixed Gaussian-Lorentzian function (GL30) following subtraction of a shirley background, as previously reported [14,15].

Results and discussion
The GIXRR measurements of Ni/Ti multilayers deposited with different R O 2 are shown as dot in figure 1. These curves associated to the interface quality exhibit an obvious difference in the cases of insufficient and abundant oxygen. Since the reflectivity profiles represent the Fourier transform of the electron density variation along the growth direction, the higher and more Bragg peaks correspond to the smoother and sharper interfaces [10]. For the multilayers deposited at and 20%, the thirteenth order Bragg peak can be evidently observed, indicating the smooth interfaces. The suppression of the even order Bragg peak intensity results from the similar individual layer thickness in the bilayers. On further increasing R O 2 to 50% and 100%, the profile of the Bragg peaks cannot be recognized after eighth order, exhibiting a striking degradation in interface quality. The GIXRR curves were fitted using Parratt's formula to obtain the key parameters including thickness and interfacial width, shown as red solid lines in figure 1. The interfacial width (s) modeled as a Gaussian distribution of the interface position is incorporated into the calculations with a Nevot-Croce factor of [ ( ) ] s --q q exp 2 , j j 1 2 / which is suitable for x-ray and neutron [16,17]. Since the Bragg peaks and the Kiessig fringes between the Bragg peaks are sensitive to the parameters including the thickness of individual layers and the interfacial width, the fitting provides well-defined values for them [18,19]. These parameters are summarized in table 1. For the multilayer deposited with 100% N 2 as reactive gas, the widths of Ti-on-Ni and Ni-on-Ti interfaces are only 0.70 nm and 0.61 nm, respectively. The two types of interfaces decrease to 0.58 nm and 0.49 nm at According to the Nevot-Croce factor, the decrease in reflectivity is exponentially dependent on the square of the interfacial width. A modest improvement in the interfacial width at can result in a remarkable increase in reflectivity performance. However, the interfacial widths dramatically increase at = R 50% O 2 and 100%. There is an indication that the optimum interface quality is obtained at The XDS measurements of Ni/Ti multilayers deposited with different R O 2 were conducted to exclusively investigate the evolution of the interfacial roughness and its nature along the whole stack, as shown in figure 2. For the multilayers deposited at and 20%, the low resonance diffusion scattering (RDS) peaks indicate a weak vertical roughness correlation. The RDS peaks disappear as the R O 2 increases to 50% and 100%, suggesting a reduction in the vertical roughness correlation.
Fittings using the distorted wave Born approximation (DWBA) quantitatively describe the characterizations of roughness structures with the lateral and vertical correlation lengths [11,20]. The fitting parameters are summarized in table 2. The vertical correlation length ξ ⊥ of the Ni/Ti multilayers decreases monotonically with the increase of R , O 2 indicating the replication of interfacial roughness can be modulated by oxygen content. It is known that roughness at the interface tends to be replicated, leading to roughness accumulation [10]. Compared to the case without O 2 , the reduction in the replication of interfacial roughness in the case with 20% O 2 contributes to suppressing the inherent roughness accumulation in Ni/Ti multilayers. With the R O 2 increasing from 0% to 50%, the lateral correlation length | x that represents the size of lateral roughness structure continuously increases from 432.37 nm to 571.37 nm, while it decreases to 486.38 nm at = R 100%.
Furthermore, the change in the fractal exponent suggests a microstructure change in the multilayers. The cross-sectional TEM images of the Ni/Ti multilayers deposited at R O 2 = 0, 20, and 50% are presented in figure 3 for an intuitive comparison. The interfacial width includes interfacial roughness and diffusion, where the interfacial roughness is the root mean square of the height deviation from its average position and the diffusion refers to the mutual penetration between two layers. In figure 3(a), the interfaces near the substrate are   figure 3(b), for the multilayer deposited at R O 2 = 20%, not only is the bottom of the multilayer smoother, but there is no significant increase in the interfacial roughness at the top of the multilayer. As the oxygen content further increases to 50%, the interfacial roughness is large even at these interfaces near the substrate, but does not significantly increase at the upper interfaces. Compared to the   figure 4. For the multilayers deposited with the low oxygen contents of 0% and 20%, the diffraction peaks can be attributed to fcc Ni (JCPDF n°04-0850) and hcp Ti (JCPDF n°89-4893). Comparing with the observed results of N-doped Ni study [21], a similar shifting towards the small angle direction and broadening of the Ni peaks were observed, indicating the lattice expansion and the fine grains. The lattice of Ni at shows a smaller degree of expansion. On further increasing R O 2 to 50% and 100%, the crystal phase transforms into fcc NiO (JCPDF n°73-1519) with the significantly narrowed diffraction peaks. The substantial transformation of the crystalline structure can cause the striking change of interfacial width observed in GIXRR. The grain size can be calculated from the full width at half maximum (FWHM) of diffraction peak using the Scherrer formula. The result presented in table 3 shows  which corresponds to the change of lateral correlation length ξ || determined by XDS. Such an association is also observed by Singh et al in their investigation into Ni/Ti multilayers on annealing [11].
For all the samples in this paper, the Ni layer hardly forms its nitride due to the chemical inactiveness between Ni and N 2 [21]. In this case, N atoms are incorporated in the interstitial sites of Ni lattice, causing the lattice expansion. Thereby, the long-range ordering is restricted and the fine grains are obtained in Ni layers. For the multilayers deposited with the oxygen content of 0% and 20%, the O-doping has the similar effect as N-doping on crystallization and the phase is attributed to Ni lattice. and 20%, the Ni 2p and Ti 2p core level spectra of the multilayers are analyzed, as performed in figures 5(a) and (b), corresponding to the positions inside Ni and Ti layers, respectively. The chemical states of elements in the two samples were found similar. The Ni 2p core level spectra exhibit that the binding energy, peak width, and satellite intensity are consistent with the spectrum of metallic Ni [23][24][25]. This agrees well with the XRD result that N or O atoms are incorporated in the interstitial sites of Ni lattice without the compound phase. Ti 2p peaks are found to be very wide and can be decomposed into the peaks of three components including Ti metal, Ti oxide (TiO), and Ti oxynitride (TiNxOy) [26][27][28]. Although high base vacuum, pre-sputtering, and residual reactive gas extraction were performed, the doping of O and N in the Ti layers is still inevitable due to the active chemical nature of Ti. The previous studies indicate that the proper doping of N or O in the Ti layers can optimize the interfaces of Ni/ Ti multilayers [22,29].

Conclusions
In this paper, the effect of the R O 2 control of reactive magnetron sputtering on the interface, microstructure, and chemical composition of the Ni/Ti multilayers has been studies. Compared to the case without O 2 , the presence of 20% O 2 decreases the average interfacial width of the Ni/Ti multilayer from 0.66 to 0.51 nm. However, the abundant O 2 substantially degrades the interface quality, leading to an increasing interfacial width to 0.82 nm. The incorporation of oxygen can suppress the vertical replication of interfacial roughness as seen from the diffuse scattering profiles, which contributes to inhabiting the inherent roughness accumulation in Ni/Ti multilayers. The striking degradation of interface quality at large oxygen content can be associated with the transformation of crystalline structure. The grain size first increases from 3.0 to 7.2 nm with the increasing oxygen content from 0% to 50%, especially in the stage of the formation of a relatively complete NiO phase with the size from 3.5 to 7.2 nm. However, the excessive oxygen decreases the grain size due to the deformation of the formed NiO phase. The change in the lateral correlation length determined by XDS analysis shows a positive correlation with the evolution of grain size. It increases slowly from 432.37 to 477.65 nm and increases rapidly to 571.37 nm with the formation of a relatively complete NiO phase. From the XPS measurements, the Ti layer exhibits the high chemical activity with N or O while the Ni layer is hard to form compound than the Ti layer. This agrees well with the XRD result that N or O atoms are incorporated in the interstitial sites of Ni lattice without the observed compound phase. This study has given us a valuable insight into the dependence of microscopic feature on the O 2 /N 2 ratio and will be helpful in the performance optimization of the multilayer device.