Introduction

Annealing studies of TiD y /Pd bi-layer films are essential to improve the performance of Ti–H (Ti–D) compounds, well-known in the application of hydrogen storage media and in catalytic and energetic reactions [1, 2], or as material for the synthesis of fine-grained Ti-based alloys [3]. Evolution of hydrogen (deuterium) from such material is realized at elevated temperatures. The results of our recent studies related to titanium-deuterated films [4] and those reported by other authors [5, 6] revealed an essential role of the thickness dependent morphology of the Ti film towards the formation of a titanium deuteride (TiD y ) film and its subsequent decomposition. Annealing-induced decomposition of fine-grained thin TiD y films of 10–20 nm thickness proceeds at much lower temperatures than the corresponding thicker films. The kinetics of this process was discussed in terms of an intermediate decomposition surface state towards the recombinative desorption of molecular deuterium [4]. In view of these results, we believe it would be interesting to investigate ultrathin TiD y /Pd bi-layer films evaporated on quartz or Si substrate where both Ti–Pd interlayer diffusion and film–substrate interaction is expected to occur as a result of annealing processing. The results of our most recent studies related to ultrathin TiD y /Pd films evaporated on quartz showed a progressive change in chemical composition within the surface and subsurface area of the film during annealing, leading to extensive inter-diffusion of Ti within the Pd top layer, strongly affecting the kinetics of deuterium desorption [7]. Si inter-diffusion from the quartz substrate layer could, however, not be observed. This observation stimulated our interest in the practical application of pure silicon as substrate for ultrathin TiD y /Pd films. Silicon plates are used very often as material support for metal film deposition, and silicon/transition metal interfaces have been intensively investigated since they enjoy wide application in highly miniaturized semiconductor devices [8].

In the present paper, we report the results of our experimental studies on the annealing of an ultrathin TiD y /Pd film evaporated on a Si(100) substrate. We want to show how annealing-induced decomposition of titanium deuteride affects both the bulk and surface structure of an ultrathin TiD y /Pd film, and what the tendency will be of the structural film as a result of annealing-induced film–silicon substrate interaction. In order to elucidate these points, we carried out a comparative investigation of Si/TiD y /Pd films taken before and after annealing. The analysis of selected samples was performed using transmission electron microscopy (TEM)—in combination with energy-filtered TEM (EFTEM) and energy dispersive X-ray spectrometry (EDX), and X-ray photoelectron spectroscopy (XPS).

Experimental

Ultrathin Si/TiD y /Pd film preparation was carried out following the same procedure as we used for the preparation of corresponding films on a quartz substrate [7]. Ten- to 20-nm-thick Ti films were evaporated within a glass UHV system [9] onto a Si(100) substrate plate kept within a glass preparation cell at 273 K at a pressure ≤1 × 10−7 Pa. After evaporation, the films were annealed for 60 min at 650 K. The TiD y films were then prepared by volumetrically controlled D2 sorption at 298 K [10] until an equilibrium pressure of approx. 1 Pa was reached. After adsorption, deuterium was evacuated to the final steady pressure approaching 10−4 Pa, at which a 10–20-nm-thick Pd layer was evaporated.

Annealing of the Si/TiD y /Pd film was performed in situ using the same procedure as we well established previously for TiD y /Pd films evaporated on quartz [7]. The procedure was determined by the experimental conditions, in which decomposition of the deuteride phase was completed. Using a heating rate of 50 K/min decomposition of titanium deuteride was accomplished in 12–15 min in a temperature range 300–673 K [7]. Deuterium evolution as a result of annealing the TiD y /Pd films was monitored in situ by mass spectrometry.

Morphological examination of the Si/TiD y /Pd films before and after annealing was performed ex situ in separate analytical systems. Therefore, in order to compare the films before and after annealing, we prepared two separate Si/TiD y /Pd film samples using the same preparation procedure. However, only one of these samples was annealed. In this way, we were able to prepare specimens representative of the Si/TiD y /Pd films prior to annealing and the corresponding films after annealing processing, respectively.

TEM analyses were carried out in a Philips CM300ST-FEG, which was equipped with a Gatan Tridiem energy filter and Thermo Fisher Noran System Six EDX analyzer with nanotrace EDX detector. The TEM specimens of the analyzed films were prepared in cross-section (XS) according to the recipe described in Ref. [11]. XS-TEM images were used to extract information regarding the bulk structure of the corresponding films. Selected area diffraction (SAD) analysis allowed the crystal phases in the bulk of Si/Pd/TiD y film to be identified. EFTEM analysis revealed the element distribution in both the bulk of films and Si substrate/film interface region. The three-window method was used for the EFTEM analysis in order to minimize interference from Pd and Ti. Verification of the correctness of the resulting Pd and Ti maps was provided by EDX analyses as well as EFTEM-spectrum imaging.

The characterization of the complex annealed TiD y /Pd films has also been carried out using XPS. The XPS spectra were recorded in a PHI 5000 VersaProbe TM spectrometer using monochromatic Al-Kα radiation (hν = 1,486.6 eV) from an X-ray source operating at 100 µm spotsize, 25 W and 15 kV. The analyzer pass energy was 23.5 eV and the energy step size was 0.1 eV. Both AR and argon-sputter profiling XPS analysis have been carried out. In the AR mode, the spectra were collected at electron take-off angles 10–90° relative to the surface plane. XPS sputter depth profiling was done using a 500-eV argon ion beam, a sputter area of 2 × 2 mm and a sputter rate of 1.4 nm/min, calibrated on a 100-nm-thick SiO2 layer thermally grown on a Si substrate. The spectra were analyzed using the Casa XPS software. Shirley background subtraction and peak fitting with Gaussian–Lorentzian-shaped profiles was performed for all considered photoelectron peaks.

Results and discussion

Cross-sectional TEM analysis

Figure 1 shows the bright-field TEM (BFTEM) cross-sectional images of an ultrathin TiD y /Pd film, recorded before (a) and after annealing-induced deuterium evolution (b). The TEM images are similar to those which we reported previously for thin TiDy/Pd films evaporated on a quartz substrate [7]. We observed distinct, sharp, and well-separated areas of the TiD y and Pd layers in the TiD y /Pd film before annealing (a), and intermixed areas of Ti and Pd films forming an amorphous bulk structure morphology as a result of annealing processing (b). The morphology of the Si material in the area of the substrate close to the annealed film interface (b) seems to be significantly affected as well: irregular changes in Si bulk morphology are clearly visible within about 15 nm depth of the substrate. Shown below the BFTEM images are the corresponding SAD patterns ((c) and (d) for the Si/TiD y /Pd film before and after annealing, respectively) from which certain crystallographic properties can be retrieved. From a direct measurement of the radius between the central spot and the diffraction spots lying on artificial concentric circles (in Fig. 1c and d shown as semitransparent gray circles drawn through the relevant diffraction spots), the Miller index planes could be determined. From the diffraction pattern describing the Si/TiD y /Pd film prior to deuterium evolution (c) the most relevant crystallographic planes belonging to the Pd fcc structure could be identified (mainly Pd (111), and in addition Pd (200) and Pd (220)). A small contribution of Ti hcp and TiSi structures should also be considered although their identification is difficult because most of the crystallographic planes belonging to Ti (Ti (101)) and TiSi (TiSi (112), TiSi (022), and TiSi (222)) overlap with the crystallographic planes of Pd [12]. No distinct diffraction spots associated with the TiD2 crystalline form were observed. The diffraction pattern of the film after annealing (d) disclosed very weak and diffuse diffraction spots, which can be assigned to coexisting Pd fcc, Ti hcp, PdTi2, and TiSi [12] crystallographic phases. However, the low-intensity spots in the diffraction pattern seem to indicate that the contribution of crystalline phases in the annealed film is very small. Supporting evidence is provided by the high-magnification cross-sectional TEM lattice images presented in Fig. 1e, f, and g: very fine polycrystalline grains (6–8 nm), originated mainly from the Pd phase in (e), are consistent with the weak diffraction spots in Fig. 1c, whereas the dominant amorphous areas in addition to the rare polycrystalline grains in the annealed film (f, g) can explain the observed weak and diffuse diffraction pattern presented in (d).

Fig. 1
figure 1

Low-magnification TEM cross-sectional bright-field images (BFTEM) of the TiD y /Pd films evaporated on Si(100) taken before (a) and after (b) annealing. The SAD patterns recorded on the corresponding films before and after annealing are shown in images c and d, respectively, together with the assignment of the most relevant low-index Miller planes of the diffraction spots. Below are shown the high-magnification cross-sectional TEM lattice images of the Si/TiD y /Pd top layer before (e) and after annealing (f, g). The Pd grains with a dominant (111) orientation are marked in (e)

Full size image

In Fig. 2, the results of the EFTEM analyses on the cross-section planes of the TiD y /Pd film taken before (left column) and after (right column) annealing-induced deuterium evolution are presented. In these columns, the BFTEM images, together with the associated EFTEM elemental mappings of Si (L2,3 edge), Ti (L2,3 edge), and Pd (M4,5 edge), are compared. The elemental distribution is marked in white color (an area with bright intensity represents a higher elemental concentration than an area with dark intensity). To visualize the elemental distribution even more clearly, the elemental maps are combined into an RGB composite image presented in the far right column of Fig. 2. In this figure, each elemental map of Si, Ti, and Pd is associated with a color, i.e., blue, green, and red, respectively.

Fig. 2
figure 2

EFTEM analysis of the Si/TiD y /Pd layer before (left column) and after annealing processing (right column). The bright-field TEM (BFTEM) cross-sectional images of both films are presented on top of each column. Below the BFTEM images are shown the associated elemental mappings of Si (L2,3 edge), Ti (L2,3 edge), and Pd (M4,5 edge). The elemental distribution is marked in white color. On the far right, the RGB images of the ultrathin Si/TiD y /Pd film taken before and after annealing are shown, as created by superimposing the elemental EFTEM maps of Si (blue), Ti (green), and Pd (red)

Full size image

The EFTEM images reveal that as a result of the annealing, an extensive intermixing process between the Ti and Pd layers occurs as well as a significant Si and Ti inter-diffusion within the Ti film and Si substrate region, respectively. Silicon is not detected on the top of annealed film but the depth of Si atom penetration within the Ti/Pd interlayer is roughly more than 15 nm as estimated from the EFTEM mapping images. One can also observe a visible enrichment of Ti on the top surface of the Pd film.

XPS analysis

The chemical nature of the components formed within the surface and subsurface areas as well as in the bulk region of the annealed TiD y /Pd films was investigated using XPS. The bulk distribution of elements within the films, taken before and after annealing, were analyzed by means of ARXPS and XPS ion argon-sputter depth profile analyses.

XPS surface and subsurface analysis

Figure 3 shows the XPS survey spectra obtained before (spectrum 1) and after (spectrum 2) annealing-induced deuterium evolution from the Pd/TiD y film. The presence of Ti peaks, in addition to those associated with Pd, can be clearly observed in spectrum 2. Silicon peaks were not detected in the surface region of both films.

Fig. 3
figure 3

Wide XPS binding energy spectra of the Si/TiD y /Pd film surface taken before (line 1) and after (line 2) annealing processing. Both XPS spectra were recorded after removing about 1.4 nm of the Pd top layer by Ar+ sputtering

Full size image

The relative distribution of Ti and Pd within the surface and subsurface region of the annealed film was revealed by the ARXPS spectra, of which the analysis results are presented in Fig. 4. This was done by measuring the areas of all detail photoelectron peaks of Pd 3d (a) and Ti 2p (b), recorded on the annealed film collected at electron take-off angles in the range of 10–90°. As a result, the relative atomic concentration ratio Ti 2p/Pd 3d at various depths within the surface and subsurface regions of the film was derived (c). The result of this analysis disclosed a significant enrichment of Ti on the top Pd layer. The chemical nature of Ti compounds at the top-most section of the film surface area can be determined from an analysis of the ARXPS spectra of Ti 2p in Fig. 4b, where at low angles (10° and 20°), only TiO2 peaks could be detected (see Ti 2p 3/2 peak at 458.6 eV, and its corresponding Ti 2p 1/2 peak at 464.2 eV BE).

Fig. 4
figure 4

Analysis of the Pd 3d (a) and Ti 2p (b) angle-resolved ARXPS spectra collected at angles in the range of 10–90° on the Si/TiD y /Pd layer after annealing: the relative atomic concentration ratio of Ti 2p/Pd 3d at various depths of the surface and subsurface regions of the annealed film is shown in (c)

Full size image

In order to analyze the chemical nature of Pd compounds and their relative depth distribution within the surface/subsurface interface, we estimated that the areas of all peaks result from a deconvolution of the Pd 3d XPS spectra. The analysis of this spectrum (Fig. 5a) reveals the coexistence of three doublets in the Pd 3d spectra (Pd 3d 5/2 peaks positioned at 335.0, 336.0, and 337.9 eV BE, and its corresponding Pd 3d 3/2 peaks at 340.3, 341.6, and 343.2 eV BE, respectively), which can be ascribed to Pd, PdTi2, and PdO2 [13]. In Fig. 5b, a relative atomic concentration depth-contribution of the detected Pd compounds within the surface–subsurface region is shown. One can see that the Pd atom concentration is relatively higher on the top film surface area than in the film subsurface and bulk region (see the low-angle XPS data). In deeper bulk film areas, the relative concentration of the PdTi2 becomes even higher than Pd (see later), which indicates that the Ti, of which the segregation was detected on the top of the Pd layer, does not only originate from the PdTi2 but also from the bottom TiD y layer forming apparently titanium oxides. This conclusion supports very well the results of the ARXPS Ti 2p measurements (Fig. 4b) presented above.

Fig. 5
figure 5

a The peak-fit analysis of the Pd 3d XPS spectrum, taken at 90° on the annealed Si/TiD y /Pd layer surface. b Analysis of the angle-resolved ARXPS spectra of the Pd 3d region. The relative atomic concentration of the Pd compounds (marked in a) within the surface–subsurface region of the annealed film was estimated from the Pd 3d XPS spectra collected at angles in the range of 10–90°

Full size image

XPS depth-profiling analysis

Both Si/TiD y /Pd samples, taken prior to and after annealing were depth-profiled using XPS analysis to reveal the Pd, Ti, and Si bulk distribution within both films. Because both films were prepared and analyzed under the same experimental conditions, we assume that all sputter-profiling-induced effects, as discussed in Ref. [14], are similar for these films. Therefore, in spite of thickness differences between the two films investigated, in a direct comparison of the elemental depth profiles, these sputter-profiling-induced artifacts are canceled out, thus disclosing only the difference in the Pd, Ti, and Si distribution in the films caused by the annealing and annealing-induced deuterium evolution processes.

Figure 6 shows the XPS depth profiles of both compared films. The depth distribution of palladium, titanium, silicon, and carbon evaluated from the Pd 3d, Ti 2p, Si 2p, and C 1s XPS peaks, respectively, is shown as a function of sputter time. The analysis of the oxygen depth distribution in the samples investigated was rather doubtful because of the overlap of the O 1s and Pd 3p 3/2 states. However, the oxygen distribution in the film–Si substrate interface region could be derived by monitoring the depth profile of SiO2, i.e., by measuring the Si 2p signal. Both samples were pre-sputtered for 1 min in order to remove the carbon surface contamination (in this time, approx. 1.4 nm surface contamination was removed).

Fig. 6
figure 6

XPS sputter depth profiles of the Si/TiD y /Pd layers taken before (a) and after (b) annealing. The relative atomic concentration depth distribution of palladium, titanium, silicon, and carbon evaluated from the Pd 3d, Ti 2p, Si 2p, and C 1s XPS peaks, respectively, are shown as a function of sputter time (sputter rate: 1.4 nm/min relative to a 100-nm-thick SiO2/Si layer). The selected Si 2p XPS profile of SiO2 is also shown for better characterization of the contaminants in the film–substrate interface area

Full size image

The elemental depth profiles for the sample prior to annealing, presented in Fig. 6a, revealed the Pd to be the sole component in the initial stage of depth profiling. At larger depths, the change from the Pd to TiD y layer and then to the Si substrate appear to proceed in a gradual fashion. The partially overlapping elemental profiles reveal an intermixing zone, which is expected to be the result of preferential ion-sputtering effects [14, 15]. More complex intermixed changes can be observed within the TiD y –Si substrate interface region where minor contamination by silicon oxide and silicon carbide was detected. The Si 2p XPS spectra recorded in depth of the Ti–Si mixed region were found to be shifted to 99.2 eV BE, lower than we determined on the pure Si substrate (99.4 eV), indicating a partial interaction of the Si substrate with evaporated Ti film leading to the formation of titanium silicide [16–18]. However, we did not observe any extensive inter-diffusion of Si through the entire set of TiD y –Pd layers.

Short-time annealing at 673 K, implicating deuterium evolution from the TiD y phase, causes substantial reaction and intermixing of Ti, Pd, and Si across the initial Pd/TiD y and TiD y /Si interfaces, as shown in Fig. 6b. As a result, the profiles for Pd, Ti, and Si are considerably broader than observed for samples prior to annealing (Fig. 6a). We observe very extensive inter-diffusion of Ti through the Pd layer and segregation towards the top Pd surface, as well as interaction of Ti with interdiffusing Si from the substrate material. Relatively broad profiles of C 1s and SiO2 were recorded within the Ti–Si interface region. However, in this case, in contrast with the unannealed film, where the C 1s spectra originated only from silicon carbide, the C 1s spectra reveal both the aliphatic carbon and silicon carbide.

From the elemental XPS spectra, recorded after the following discrete steps during depth profiling, various chemical compounds formed at various depths of the annealed film could be identified. As an example, in Fig. 7, we present the result of an analysis of the Si 2p profile. The chemical nature of these spectra is illustrated in Fig. 7a, in which also a Si 2p spectrum is shown, which was recorded in a separate experiment from pure substrate material (spectrum (e)). One can see that all Si 2p spectra, recorded in the depth profiling of the annealed sample, are shifted towards lower BE values as compared with the pure Si material, an indication that titanium silicide is the dominant intermetallic compound in the whole analyzed depth area of the film. The relatively high shift in BE (−0.6 to −0.9 eV) is more likely to be associated with TiSi (−0.6 eV [17]) than with TiSi2 (−0.2 eV [18]). Therefore, for the analysis of the Si 2p depth profile, we used the BE value of our pure reference Si substrate (99.4 eV) correlated with BE values found by other authors for TiSi (BE shift −0.6 eV [17]) and Si in solution in Ti (BE shift −1.1 eV [17]): The coexistence of silicon oxides and palladium silicides was also analyzed although the Si 2p peak in PdSi shows significant overlap with that in Si (Si 2p BE 99.5 eV [19] and 99.4 eV [16], respectively). As shown in Fig. 7b, TiSi and interdiffused Si are major components within the annealed film. On the other hand, titanium silicide is also a dominant intermetallic component found in the top of the substrate layer. This result is in agreement with the EFTEM images reflecting heavy Ti inter-diffusion within both the Pd film and Si substrate phases. Silicon oxide has only been observed within the film–substrate interface.

Fig. 7
figure 7

a The Si 2p XPS spectra taken after subsequent steps of Ar+ sputtering (see lines marked as a, b, c, and d). The Si2p spectrum (e) was recorded separately from pure Si(100) substrate material. b The analysis of the Si 2p XPS sputter depth profile of the annealed Si/TiD y /Pd layer. For a detailed description, see the text

Full size image

Discussion of annealing processes

The experimental data presented above shows the effects of two various/subsequent annealing procedures applied in the Si/TiD y /Pd film processing. The first annealing procedure was carried out routinely for both compared samples as part of one of the stages of the Ti film preparation prior to deuterium sorption. The aim of this annealing stage was to remove all absorbed gasses (mainly nitrogen) from the Ti film, which could affect the subsequent deuterium sorption. In our case, the effects of this annealing processing step can be analyzed in the “as prepared” Si/TiD y /Pd film. Taking into account the time and temperature of this annealing step (60 min, 650 K), an initial Ti–Si reaction is expected to occur leading to the observed TiSi and TiSi2 phase formation. Similar low-temperature (approx. 570 K) initial reactions at the Ti–Si interface has been reported by other authors [15, 17]. Both the SAD patterns presented in Chapter 3.1 (Fig. 1 c and d) and the Si 2p XPS spectra provided evidence for the coexistence of a titanium silicide phase in the “as prepared” Si/TiD y /Pd film. However, sharp boundaries between the PdD y , Ti, and Si phases in the EFTEM images (Fig. 2) indicate that there is no significant amount of inter-diffusion of Si. It is likely that inter-diffusion of Si into Ti is partially retarded by SiO2 and SiC species [20], whose small contamination level has been detected in the film–substrate interface (see XPS depth profiles in Fig. 6a).

The second annealing procedure we used for the finally prepared Si/TiD y /Pd film in order to evolve deuterium from the TiD y phase. This process was completed in 12–15 min in the temperature range 300–675 K using a heating rate of 50 K/min. In spite of the fact that the final annealing temperature was only 25 K higher than used for the preliminary Ti/Si film annealing, and the associated time of annealing being much shorter, we observed a substantial structural and chemical rearrangement of the film. Both EFTEM images and XPS sputter profiles provided evidence for an extensive inter-diffusion of Ti, Pd, and Si. Similar annealing-induced inter-diffusion phenomena of Ti and Pd have been reported previously for thin TiD y /Pd films prepared on quartz [7]. However, we did not find any evidence for serious inter-diffusion effects of Si from the quartz substrate layer. In contrast, the annealing of the ultrathin TiD y /Pd films prepared on a Si(100) substrate induced an extensive interaction of Si with these films.

This work shows that interaction with deuterium atoms, released during annealing, induces various processes leading to a chemical and structural transformation of the Si material within the film–substrate interface and accelerating reactive inter-diffusion of Si and Ti. Hydrogen (deuterium) interaction with Si has been the subject of active research reported well in literature (see, e.g., Ref. [21]). Low-temperature hydrogen interaction with pure Si material was found to form SiH x species, which decomposes at temperatures higher than 670 K [21, 22]. However in our films, this ideal scheme of deuterium–silicon interaction should be considered as just one of various processes proceeding simultaneously in the TiD y –Si interface. Indeed, deuterium released from the TiD y phase can interact with the Si substrate interface forming Si–D species and initialize interaction of pure Ti decomposed from the TiD y phase with Si. This simple mechanism is however, complicated by SiO2 and TiO x interface contaminants and also by the structural phase transformation effects like amorphization or stress. The formation of amorphous material is usually observed as an intermediate state of crystal phase transformation. This tendency is well supported by the BFTEM cross-section and SAD images of the annealed film (see Fig. 1). Stress accompanying the structural phase transformation was considered by other authors [23] as a source of the very early stages of the Ti–Si reaction and TiSi2 phase growth. On the other hand, tensile stress can induce deformation of the film and its separation from the substrate. We observed these two phenomena on our annealed samples. The BFTEM images reveal a visible amorphous Ti–Si interlayer (see Fig. 1), EFTEM shows extensive penetration of Ti within the Si (Fig. 2) and XPS depth profile analysis discloses much higher concentrations of SiC, TiO2, and SiO2 compared with the “as prepared” sample (Fig. 6). These observations, as well as the detection of aliphatic carbon in the interface region indicate a partial detachment of the film–substrate interlayer as a result of annealing.

Summary and conclusions

Structural and chemical phenomena accompanying the annealing of ultrathin TiD y /Pd films evaporated on a Si(100) substrate have been investigated. It was found that the short annealing time at 673 K, inducing deuterium evolution from the TiD y phase, causes substantial reactions and intermixing of Ti, Pd, and Si across the initial Pd/TiD y and TiD y /Si interfaces. The surface and bulk morphology of the bi-layer film as well as the Si substrate material undergo microstructural and chemical conversions. Broad intermixing between the Ti and Pd layers and an extensive inter-diffusion of Si and Ti within the film–substrate interlayer have been observed as evidenced by energy-filtered TEM and XPS argon ion sputter depth profiling. The cross-section TEM images reveal a dominant contribution of the amorphized material in the annealed film. However, SAD pattern analysis revealed also the presence of very fine crystallites of PdTi2 and an indication of the formation of a TiSi phase in the initial stages. Evidence for segregation of Ti towards the Pd top layer surface was provided by angle-resolved XPS and EFTEM.